Furnace Core

NUKEX USER GUIDE

VERSION 4.3V4

©2012 The Foundry Visionmongers Ltd. All rights reserved. FurnaceCore User Guide

This manual, as well as the software described in it, is furnished under license and may only be used or copied in accordance with the terms of

such license. This manual is provided for informational use only and is subject to change without notice. The Foundry assumes no responsibility or liability for any errors or inaccuracies that may appear in this book.

No part of this manual may be reproduced, stored in a retrieval system, or transmitted in any form without the prior written permission of The Foundry.

The Foundry logo is a trademark of The Foundry Visionmongers Ltd. Nuke is a registered trademark of The Foundry Visionmongers Ltd. All other

products or brands are trademarks or registered trademarks of their respective companies or organisations.

The Foundry algorithms use the FFTW library developed by Matteo Frigo and Steven G. Johnson, copyright 2003 Matteo Frigo, copyright 2003

Massachusetts Institute of Technology. All rights reserved. Used under terms of a commercial license. http://www.fftw.org.

Software engineering: Simon Robinson, Ben Kent, Ralph McEntagart, Lucy Wilkes, Phil Parsonage, Andy Whitmore, Bruno Nicoletti, and Mailys

Levassort.

Algorithms: Dr. Bill Collis, Dr. Anil Kokaram of Trinity College Dublin, Prof. Paul White of the University of Southampton, Ben Kent, Phil Parsonage,

and Dr. Francois Pitie.

Product testing: Jack Binks, Dan Allum, Joel Braham, and Antony Nasce.

Writing and layout design: Eija Narvanen using FrameMaker.

Proofreading Eija Narvanen.

The Foundry FurnaceCore for NukeX

Contents

Introduction About this User Guide. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

OFX Plug-ins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

What’s New? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Installing FurnaceCore . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Licensing FurnaceCore . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

On-Screen Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Motion Vector Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Colour Space in FurnaceCore Plug-ins . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

Customer Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

Other Foundry Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

Overview Grain Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

Retiming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Removing Flicker and In-Scene Objects . . . . . . . . . . . . . . . . . . . . . . . . . 11

Stabilisation and Alignment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

Grading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

Align Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Quick Start . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

DeFlicker2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

Quick Start . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

Kronos Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

Quick Start . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

MatchGrade Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

Quick Start . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

4

Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

MotionBlur Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

Quick Start . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

ReGrain Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

Colour Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

Quick Start . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

Response. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

Checking the Result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

Proxy Resolutions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

RigRemoval Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

Quick Start . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

Steadiness Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

Quick Start . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

VectorGenerator Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

Output. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

Quick Start . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

WireRemoval Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

Quick Start . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

Indicators on the On-Screen Wire Tool . . . . . . . . . . . . . . . . . . . . . . . . . 65

Tracker Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

FurnaceCore for NukeXThe Foundry

5

Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

Global Motion Estimation

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

What is Global Motion Estimation? . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

The Analysing Global Motion Estimation Effects . . . . . . . . . . . . . . . . . . 72

Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

Widgets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

Local Motion Estimation

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

Using Pre-Calculated Vector Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

Vector Field Representation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

Appendix A Third Party Licenses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

Appendix B End User Licensing Agreement (EULA) . . . . . . . . . . . . . . . . . . . . . . . . . . 84

Index A-Z . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91


INTRODUCTION

Welcome to this User Guide for FurnaceCore on NukeX.

About this User Guide

This user guide describes FurnaceCore - the twelve most popular Furnace plug-ins incorporated in NukeX.

Throughout this user guide, we assume you are familiar with the version of NukeX you are running.

OFX Plug-ins These plug-ins have been written and compiled to the OFX plug-in standard. OFX is an open plug-in API for 2D visual effects. For a current list of supported host systems see http://www.thefoundry.co.uk.

What’s New? For new features and feature improvements, have look at the Nuke Release Notes (in Nuke, click Help > Release Notes).

Notation In this user guide, we will refer to machines running FurnaceCore and NukeX as clients and machines that are running the Foundry FLEXlm Tools as servers.

Installing FurnaceCore

FurnaceCore is installed automatically with the Nuke installation package, available as a download from our web site, http://www.thefoundry.co.uk. The downloads are in compressed tar format (tgz) for Linux, package format (dmg) for Mac OS X, and executable installer format for Windows machines. For more information, see the Nuke User Guide.

Default Install Directory

By default, FurnaceCore gets installed into the following location.

On Windows: C:\Program Files\Nuke[version number]\plugins\ orC:\Program Files (x86)\Nuke[version number]\plugins\

On Linux:

INTRODUCTION 7Licensing FurnaceCore

/usr/local/Nuke[version number]/plugins

On Mac OS X: /Applications/Nuke[version number]/Nuke[version number].app/Contents/

MacOS/plugins/

Licensing FurnaceCore

To run the FurnaceCore plug-ins included in NukeX, you need a valid license key for NukeX. The license key is a sequence of numbers and letters, stored in a plain text file, that unlocks the software. For more information on licensing Nuke and NukeX, see the Nuke user guide.

Tools to install license keys, manage floating licenses, and diagnose license problems can be downloaded from our web site, http://www.thefoundry.co.uk/licensing.

On-Screen Tools Some plug-ins have their own on-screen tools for region selection or to facilitate the changing of various parameters.

All on-screen tools are transparent when not active. Parts of the widget that can be moved become solid when the mouse is moved near to them. When selected, they are extended out to the edges of the viewing area.

The rectangular tools used to select a region - for sampling or analysis, for example - sometimes apply to a single frame only. In this case, the on-screen tool will have continuous lines on this frame, and dotted lines on all other frames. When a region is selected, the frame associated with it will be automatically updated to the current frame.

Motion Vector Inputs

Many FurnaceCore plug-ins, such as F_Kronos, rely on Local Motion Estimation, which is a per-pixel analysis of the motion between pairs of frames. Several of these plug-ins, including F_Kronos, now have optional motion vector inputs, to allow you to reuse the results from previous analyses of the motion. This is designed to save processing time, as local motion estimation is computationally intensive, so it makes sense to avoid doing these calculations more than once if possible. The F_VectorGenerator plug-in allows you to do local motion estimation in isolation, so that the results can then be passed to other FurnaceCore effects.


INTRODUCTION 8Colour Space in FurnaceCore Plug-ins

Colour Space in FurnaceCore Plug-ins

Some of the algorithms in the FurnaceCore tool set are sensitive to the colour space of the source footage. If the footage is not in the expected colour space, you may get poor results from some of the plug-ins.

FurnaceCore on NukeX expects all footage to be in Linear space.

NukeX converts all footage to Linear upon import, so unless you have changed the colour space in your Read nodes (or in the node tree by using a Colorspace node before a FurnaceCore plug-in), your footage should be Linear by the time it reaches the plug-in anyway.

Plug-ins most sensitive to colour space are:

• F_MatchGrade

• F_ReGrain

If you know the input to one of these colour-space sensitive plug-ins isn’t Linear, you should use a Colorspace node before and after the plug-in to convert to and from Linear for processing. This should ensure optimal results.

F_ReGrain also has additional pop-up menus to allow you to adjust the algorithm to get the best results when using footage that was originally in another colour space. Please take a moment to read through the relevant chapter in this manual for this plug-in, to make sure you are aware of how F_ReGrain behaves.

Customer Support Should questions arise that this user guide fails to address, you can contact Customer Support directly via e-mail at [email protected] or via telephone to our London office on +44 (0)20 7968 6828 or to our Los Angeles office on (310) 399 4555 during office hours.

Other Foundry Products

The Foundry is a leading developer of plug-in visual effects for film and video post production. Its products include Nuke, a high-end compositing application, and plug-ins, such as Furnace, Keylight, Ocula, and RollingShutter. The plug-ins run on a variety of compositing platforms, including After Effects, Autodesk® Media and Entertainment Systems, Avid DS, Nuke, Shake, and Final Cut Pro. For the full list of products and supported platforms see our web site http://www.thefoundry.co.uk.

Nuke is an Academy Award® winning compositor. It has been used to create extraordinary images on scores of feature films including Australia,


INTRODUCTION 9Other Foundry Products

The Dark Knight, The Golden Compass, Iron Man, Transformers, King Kong, and Pirates of the Caribbean: At World’s End.

Keylight is an industry-proven blue/green screen keyer, giving results that look photographed, not composited. The Keylight algorithm was developed by the Computer Film Company who were honoured with a technical achievement award for digital compositing from the Academy of Motion Picture Arts and Sciences.

Ocula is a collection of tools that solve common problems with stereoscopic imagery, improve productivity in post production, and ultimately help to deliver a more rewarding 3D-stereo viewing experience.

RollingShutter is a plug-in that tackles image-distortion problems often experienced by users of CMOS cameras. The plug-in will often vastly improve the look of distorted footage, by either minimising or eradicating image distortions. Unlike solutions tied to camera stabilisation, that stretch the image as a whole, the RollingShutter plug-in compensates for local skewing and distortion in the scene, by correcting each object individually.

Visit The Foundry’s web site at http://www.thefoundry.co.uk for further details.


OVERVIEW

Great - so you’ve just downloaded, installed, and licensed up your copy of FurnaceCore, but where do you start? In this section, we’ll have a quick look at the plug-ins, broken down by what we would envision their application to be. Please bear in mind, many of the tools offer much greater flexibility than we can cover in this manual, so have a good play around with them.

Please select an overview from the list of sections below. When you see an individual plug-in that you’d like to find out a little more about, click on it and you will be taken to the plug-in reference section for that tool. Grain Management | Retiming | Removing Flicker and In-Scene Objects | Stabilisation and Alignment | Grading.

Grain Management The FurnaceCore grain management tools allow you to quickly replicate grain patterns from both modern day fine-grained film stock and old, degraded, archive material. So where (and why) would you want to manage grain? In a whole host of different places; for example:

• When comping (or even editing) two shots together it is very obvious to the viewer when the grain characteristics vary; take, for example, comping a moving Computer Generated (CG) element against a still frame: the background still has static, unmoving grain, and the CG element will have no grain at all. In this circumstance, you’d replicate the grain from the still frame and generate further frames so that the grain appears to be in motion. This would then be applied back against both the original still and the CG element, helping create the illusion that it was all shot in camera.

ReGrain allows you to reapply grain to a source clip. A number of preset stocks are included, or you can sample a plate of your choice and have ReGrain create a statistically identical moving grain sample.

Figure 1. ReGrain tool

OVERVIEW 11Retiming

Retiming Retiming is the process of altering the speed of a clip, to get either slow or fast motion. Historically, this was an ugly process which introduced juddery or artefact-ridden output. By making a best-guess about where objects move within a scene (motion estimation), we are able to create frames between existing ones. Once we know how these objects move, we are also able to add motion blur to these objects in motion as though it was shot in-camera.

Kronos is FurnaceCore’s retimer. It gives you full control over the speed of playback of your clip and allows you to add motion blur.

MotionBlur allows you to add true-to-life motion blur to objects within a scene, without the added complexity of the clip retiming controls.

Removing Flicker and In-Scene Objects

DeFlicker2 deals with film flicker from a variety of sources. It works particularly well on in-scene flicker, poorly synchronised light rigs, and stray light.

RigRemoval Got an object moving in relation to the background (or vice versa), that you want to get rid of? RigRemoval scans forwards and backwards within a sequence to find an area where the background in question was unobstructed by the object, and then copies that back into place. Great for unwanted traffic, people, and more.

WireRemoval That perennial paint favourite; cloning out, frame-by-frame, the wires used for death defying stunts. Not anymore! WireRemoval has a variety of different repair types to remove the wire automatically.

Figure 2. Retiming tools.

Figure 3. Clean-up tools.


OVERVIEW 12Stabilisation and Alignment

Stabilisation and Alignment

So you’ve got something in a wobbly shot you want to repair? These tools are the answer.

Steadiness Suppress wobbles in handheld footage or lock the sequence position against a certain frame (great for when a small shake hits an otherwise locked off shot), Steadiness requires no tracking markers to be specified.

Align Lock one shot of a scene against another. Handy for anything from doubling up the crowd size by comping together two shots, to locking your freshly generated clean plate to the original.

Grading Help speed up the grading process using this image-enhancing tool.

MatchGrade Got two shots from different times of day that need to be comped or edited together? This plug-in analyses the colour histogram of a reference image and automatically applies the result to your source sequence. This allows sequences that were shot with subtly different lighting to have the same ’look’.

Figure 4. F_MatchGrade.


ALIGN

This chapter looks at how to use F_Align to register (line up) two shots that are of the same scene, but have slightly different camera motion and foreground objects. This can be useful, for example, for doubling up the crowd size by lining up and comping together two shots of the same scene, or locking your freshly generated clean plate to the original.

Introduction F_Align takes two sequences that were shot of the same scene and lines them up spatially. It uses Global Motion Estimation (GME) to calculate a four-corner pin so that each frame in one shot (the source input) will be aligned with the corresponding frame in a second shot (the reference input). The result is the source image which has been repositioned to line up with the reference image.

To be able to align the sequences, F_Align analyses them for global motion. This analysis can be triggered for the complete sequence, specified frame

Figure 5. Source image. Notice the position of the background.

Figure 6. Reference image.

Figure 7. Output from F_Align. The source image has been

repositioned so that the background lines up with the

reference image.

Figure 8. The output of F_align comped together with the

reference image.

ALIGN 14Quick Start

range, or a single frame when you press the Analyse button in the F_Align controls. Alternatively, it can be done on the fly for a single frame when you move to a new frame on the timeline. The advantage of pressing Analyse is that during the analysis, F_Align stores the calculated four-corner pin as key-framed parameters. When you then render the output of the plug-in later, F_Align can use these key frames without having to calculate them again.

If you analyse on the fly, you won’t have direct access to the calculated corner pins. Any re-rendering will also be significantly slower, as the ’on the fly’ calculations will have been lost and F_Align will have to analyse again.

If at any stage you modify the effect in such a way to invalidate the key-framed analysis (for example by changing the Accuracy parameter), a warning will be posted and the effect will analyse on-the-fly during render, ignoring the keyed analysis.

The on-screen widget and the Analysis Region parameters are used to control which section of the Reference frame is being matched to each Source frame. Typically, leaving the region at its default is good enough. However, a heavy mismatch in foreground detail may make it necessary to change the region to a section that is shared between shots.

The transformation in F_Align is concatenated with other NukeX transform nodes. This means that if you add a row of F_Align and NukeX transform nodes to a tree, their functions are combined. Because the image is only resampled once, there is no loss of image quality and processing time is decreased. However, as in NukeX, certain nodes, including colour correction nodes, can break the concatenation.

For more of an overview of Global Motion Effects, and a description of the common way of working many of these effects have, please see the Global Motion Estimation chapter on page 71.

Quick Start This section gives a very brief outline of how to use the plug-in. It covers both analysing using the Analyse button and analysing on the fly.

Analysing with the Analyse button

To align two shots and store the results as key-framed parameters, do the following:

FurnaceCore on NukeXThe Foundry

ALIGN 15Quick Start

1. Find two shots that are of the same scene, but have slightly different camera motion and foreground objects. Select Image > Read to load both these shots.

2. Select Furnace > F_Align. Connect the shot you want to reposition to the source (Src) input of F_Align and the shot you want to match to the reference (Ref) input. View the output from F_Align.

The source will be immediately repositioned so that it aligns to the reference shot without any need for analysis.

You will see the following banner in the overlay:

Note: no valid key framed analysis found. Analysing during render.

3. Depending on the exact difference between the two shots, you may need to enable the Scale and/or the Perspective toggles to get a decent alignment.

4. You may also need to reposition the Analysis Region depending on the differences in foreground detail. However, leaving it at the default position works well for most shots.

5. Click on the Analyse button.

F_Align will now start analysing each frame in the shot, figuring out the four-corner pin and writing it as key frames to the corner pin parameters, Bottom Left, Bottom Right, Top Left, and Top Right. You will find these parameters under Advanced > Four Corner Pin.

F_Align will update the timeline at each frame, and you will see the aligned image render in the output. If you don’t want this to happen, uncheck Render During Analysis.

If you interrupt the analysis, the pins it has keyed until that point will be retained.

6. Play or scrub through the aligned frames. The rendering will be faster as F_Align will no longer need to analyse on the fly. However, if you scrub to a frame where a corner pin has not been keyed, F_Align will re-analyse that frame on the fly.

7. To see how closely the two clips have aligned, press M on the Node Graph to insert a Merge node. Connect the Merge node’s A input to F_Align and the B input to the reference image. View the output from the Merge node. Then, adjust the mix slider in the Merge controls to dissolve between F_Align’s output and the reference clip you wanted to match.

Analysing On The Fly To align two shots and calculate the alignment on the fly, do the following:

1. Find two shots that are of the same scene, but have slightly different camera motion and foreground objects. Select Image > Read to load both these shots.

2. Select Furnace > F_Align.


ALIGN 16Inputs

3. Connect one of the shots to the source (Src) input of F_Align and the other to the reference (Ref) input. View the output from F_Align.

The Src will be immediately repositioned so that it aligns to the Reference shot without any need for analysis.


Note: no valid key framed analysis found. Analysing during render.

4. Depending on the exact difference between the two shots, you may need to enable the Scale and/or the Perspective toggles to get a decent alignment.

5. You may also need to reposition the Analysis Region depending on the differences in foreground detail. However, leaving it at the default position works well for most shots.

6. To see how closely the two clips have aligned, press M on the Node Graph to insert a Merge node. Connect the Merge node’s A input to F_Align and the B input to the reference image. View the output from the Merge node. Then, adjust the mix slider in the Merge controls to dissolve between F_Align’s output and the reference clip you wanted to match.

Inputs F_Align has two input clips: the source (Src) input is moved so that each frame matches the reference (Ref) input frame.

Parameters The parameters for this plug-in are described below.

Analyse - This is a push button which will trigger an analysis of the input clips and calculate a corner pin. Interrupting the analysis will not delete the corner pin keys that have already been calculated.

Render During Analysis - If set, this toggle will cause the effect to update the time line and render a freshly analysed frame in the Viewer so you can see the progress of the effect. Doing so will slow down the analysis somewhat, so toggle this off to speed up the general analysis.

Clear Analysis - Pressing this push button will delete all key frames from the corner pin parameters, allowing you to force a re-analysis if you feel the need to.

Analysis Range - This controls the range of frames any analysis will be run over. It can be one of:

• Specified Range - which will look at the parameters Analysis Start and Analysis Stop for the range of frames to analyse,


ALIGN 17Parameters

• Source Clip Range - which will automatically determine the range of frames to analyse from the length of the input clip.

• Current Frame - the analysis occurs only on the current frame. This is useful for correcting any errors that may have occurred while analysing the entire clip.

Analysis Start - The first frame to analyse from if Analysis Range is set to Specified Range.

Analysis Stop - The last frame to analyse from if Analysis Range is set to Specified Range.

Scale - A toggle that indicates whether the calculated corner pin can include a scaling factor.

Rotate - A toggle that indicates whether the calculated corner pin can include rotations.

Translate - A toggle that indicates whether the calculated corner pin can include translations in x and y.

Perspective - A toggle that indicates whether the calculated corner pin can include perspective transforms.

Analysis Region - This is the region analysed to calculate the four-corner pin. This is especially useful when doing any form of frame locking, in which case, go to the lock frame, look at the reference clip and position the box over the area you want locked.

Analysis Region BL - controls the position of the bottom left corner of the analysis region.

Analysis Region TR - controls the position of the top right corner of the analysis region.

Advanced - The lesser used refinement controls.

Accuracy - This controls the time/accuracy trade off. The higher this is, the slower the analysis, but you have a better likelihood of a good result.

Filtering - Sets the filtering quality.

• Low - low quality but quick to render.

• Medium - uses a bilinear filter. This gives good results and is quicker to render than high filtering.

• High - uses a sinc filter to interpolate pixels giving a sharper repair. This gives the best results but takes longer to process.


ALIGN 18Parameters

Invert - if set, then the inverse of the calculated four-corner pin is used during render.

Four Corner Pin - The corner pins calculated during the analysis pass:

Bottom Left - the lower left corner pin.

Bottom Right - the lower right corner pin.

Top Left - the upper left corner pin.

Top Right - the upper right corner pin.


DEFLICKER2

When working in film, you sometimes have to deal with shots that have a luminance flicker. This chapter concentrates on removing flicker using F_DeFlicker2.

Introduction F_Deflicker2 is used to remove flicker. It is particularly suitable for removing flicker that is localised and dependent on the geometry of the scene (that is, flicker that is not present across the whole of the image), such as that caused by an unsynchronised fluorescent light in a shot. It works by calculating the gain between the current frame and each frame in a small analysis range surrounding it. It then tries to adjust the gain so that it varies smoothly over this frame range. This means it is better at reducing fast flicker than flicker which varies slowly over the image sequence, as the latter will already appear smooth over the small frame range and F_DeFlicker2 will leave it largely untouched.

The algorithm used by F_DeFlicker2 can introduce blurring in areas where there is rapid motion. This problem could be alleviated by using local motion estimation before deflickering the frames. However, this process is complicated by the fact that the presence of flicker can adversely affect the results of the motion estimation. F_DeFlicker2 therefore adopts a two-stage approach to this problem. First, the normal deflickering process is performed. Then, the motion vectors for the sequence are calculated on the resulting deflickered frames, and applied to the original frames in order to align them. The deflicker calculation is then performed on the aligned frames to give the final result. To use this approach, turn on Use Motion in F_DeFlicker2.

Note that because F_DeFlicker2 looks at input frames outside the current frame when performing its calculation, it can be a computationally expensive plug-in. As such, using more than two instances of F_DeFlicker2 in a node tree will dramatically increase render times. It is strongly advised therefore, that you render each instance out separately.

Quick Start To remove flicker from a sequence:

1. Select Image > Read to load the sequence you want to remove flicker from.

2. Select Furnace > DeFlicker2 to apply DeFlicker2. View its output.

DEFLICKER2 20Inputs

3. If you’re not happy with the results, adjust the DeFlicker2 parameters. The available parameters are described below.

4. Select Image > Write to insert a Write node and render the result.

Inputs F_DeFlicker2 has a single input: the sequence to deflicker.


DeFlicker Amount - Use this to reduce flicker without removing it entirely; smaller values mean more will be left behind.

Block Size - To find where a certain pixel is located in the analysis range, the deflicker algorithm looks for a block of pixels centered around that pixel. Block size defines the width and height of these blocks (in pixels). On rare occasions, a large block size can produce data that’s lacking in detail. This is because a small feature can fit into a large block, causing the motion estimation to concentrate on the background motion and ignore the small feature. A small value, instead, can produce a noisy motion field, as there aren’t enough constraints in a small block to fit the motion accurately. In most cases, however, the default value is small enough so that details aren’t lost, and the smoothing step of the algorithm ensures the motion field isn’t too noisy. Therefore, this value very rarely needs editing.

Use Motion - Turn this on to do a second deflicker pass using motion-compensated frames. This can improve results in areas where there is fast motion, where the initial deflicker pass can introduce blurring.

Vector Detail - Determines the density of the motion vectors used when Use Motion is turned on. The maximum value of 1 will generate one vector per pixel. This will produce the most accurate vectors but will take longer to render. A value of 0.5 will generate a vector at every other pixel.

Analysis Range - The number of frames searched each side of the current frame when calculating the flicker. Higher values may give better results, but can also bring in erroneous information and take longer to process.


KRONOS

This chapter looks at retiming sequences using F_Kronos.

Introduction F_Kronos is FurnaceCore’s re-timer and is designed to slow down or speed up footage. It works by calculating the motion in the sequence in order to generate motion vectors. These motion vectors describe how each pixel moves from frame to frame. With accurate motion vectors, it is possible to generate an output image at any point in time throughout the sequence by interpolating along the direction of the motion.

By default, F_Kronos is set to perform a half speed slow down. This is achieved by generating a new frame at position 0.25 and 0.75 between the original frames at 0 and 1. Frames are created at a quarter and three quarters instead of zero (an original frame) and a half so as not to include any original frames in the re-timed sequence. This avoids the pulsing that would otherwise be seen on every other frame on a half speed slowdown.

When F_Kronos is used to speed up motion, motion blur will be seen.

F_Kronos contains a number of controls to allow you to trade off render time versus accuracy of vectors. For more information on vectors and local motion estimation in general, “Local Motion Estimation” on page 76.

Quick Start To retime a sequence, do the following:

1. Select Image > Read to load the sequence you wish to retime.

2. Select Furnace > F_Kronos to apply F_Kronos and view the output.

By default, the Speed control will be set to perform a half speed slow down.

Figure 9. Simple mix of two frames to achieve an inbetween

frame.

Figure 10. Kronos vector interpolation of the same two

frames.

KRONOS 22Inputs

3. To adjust the slow down or to speed the sequence up, enter a new value for the Speed control. Values below 1 slow down the clip. Values above 1 speed up movement. The default value, 0.5, created the half speed slow down. Quarter speed would be 0.25.

Alternatively, you can describe the retiming in terms of ‘at frame 100 in the output clip, I want to see frame 50 of the source clip‘. To do so, set Timing to Source Frame. Go to a frame in the timeline, and set Frame to the input frame you want to appear at that output position. You’ll need to set at least two key frames for this to retime the clip. For example, to slow down a 50 frame clip by half, you can set Frame to 1 at frame 1, and to 50 at frame 100.

4. To create an arbitrarily changing speed for the sequence, see “Varying the Speed” below.

5. Render the retimed sequence using the default parameter settings. If you are not happy with the results, adjust the Vector Generation Parameters described under “Local Motion Estimation” on page 76.

Varying the Speed To vary the speed in your sequence, do the following:

1. In the F_Kronos controls, choose Source Frame as the Timing method. This allows you to describe the retiming in terms of ‘at frame 100 in the output clip, I want to see frame 50 of the source clip‘.

2. Animate the Frame parameter. Select an output frame from the timeline and set Frame to the input frame you want to appear at that output position. From the animation menu next to the Frame parameter, select Set key. Move to another frame on the timeline and set Frame to the input frame you want to appear at that position. A key frame is set automatically. For example, to do a four times slow down, move to frame 1 and set Frame to 1, select Set key from the animation menu, move to frame 19, and set Frame to 5.

You should now have a linear time curve, which you can see if you select Curve editor from the animation menu next to Frame. By using the Curve Editor to adjust this curve, you can create an arbitrarily changing speed for the sequence.

Instead of animating the Frame parameter, you can also switch Timing to Speed, and animate the Speed parameter.

Inputs F_Kronos has the following inputs:

• Source (Src) - The source image to retime.


KRONOS 23Parameters

• Motion Source (MoSrc) - If supplied, motion vectors will be calculated from this sequence and applied to the input sequence. This can be useful if, for example, your input sequence is very noisy, as too much noise interferes with the motion estimation. In that case, you should supply a smoothed version of the sequence here.

• Matte - This sequence will be used as a foreground matte. This can improve the motion estimation by reducing the dragging of pixels that can occur between foreground and background objects.

• Background Vectors (BgVec) and Foreground Vectors (FgVec) - If the motion in your input sequence has been estimated before (for example, using F_VectorGenerator) and you have the motion vectors available, you can supply one or more vector sequences to F_Kronos. This will save processing time, as F_Kronos will not then have to perform the motion estimation a second time.

If your vectors were calculated using a foreground matte, you should supply the background and foreground vector sequences as Background Vectors and Foreground Vectors respectively. In this case, you should also supply the foreground matte sequence used to create the vectors as the Matte input.

If you have a single set of vectors, you can connect it into either the Background Vectors or Foreground Vectors input.

If you have no precalculated vectors, you can ignore the vector inputs, and let F_Kronos generate the motion vectors.


Method - Sets the interpolation algorithm.

• Nearest Frame - the nearest original frame is displayed.

• Frame Blending - a mix between two frames is used for the inbetween frame. This is quick to render and useful when tweaking the timing on the curve before setting Method to Motion Estimation.

• Motion Estimation - vector interpolation is used to calculate the in-between frame.

Timing - Sets how to control the new timing of the clip.

• Speed - select this if you wish to describe the retiming in terms of overall duration, i.e. double speed will halve the duration of the clip or half speed will double the duration of the clip.

• Source Frame - select this if you wish to describe the retiming in terms of ‘at frame 100 in the output clip, I want to see frame 50 of the source clip‘. You’ll need to set at least 2 key frames for this to retime the clip.


KRONOS 24Parameters

Speed - This parameter is only active if Timing is set to Speed. Values below 1 slow down the clip. Values above 1 speed up movement. For example, to slow down the clip by a factor of two (half speed), set this value to 0.5. Quarter speed would be 0.25.

Frame - This parameter is active only if Timing is set to Source Frame. Use this to specify the source frame at the current frame in the timeline. For example, to slow down a 50 frame clip by half set the Source Frame to 1 at frame 1 and the Source Frame to 50 at frame 100. The default expression will result in a half-speed retime.

Vector Detail - Adjust this to vary the density of the vector field. The larger vector detail is, the greater the processing time, but the more detailed the vectors should be. A value of 1.0 will generate a vector at each pixel. A value of 0.5 will generate a vector at every other pixel. For some sequences, a high vector detail near 1.0 generates too much unwanted local motion detail, and often a low value is more appropriate.

Smoothness - Vector fields usually have two important qualities: they should accurately match similar pixels in one image to another and they should be smooth rather than noisy. Often, it is necessary to trade one of these qualities off against the other. A high smoothness will miss lots of local detail, but is less likely to provide you with the odd spurious vector (think of it as if the vectors were combed until they are smooth with respect to their neighbours). A low smoothness will concentrate on detail matching, even if the resulting field is jagged. The default value of 0.5 should work well for most sequences.

Shutter - Controls for adjusting the shutter time and the smoothness of the motion blur.

Shutter Time - Sets the equivalent shutter time of the retimed sequence. A shutter time of 1 is equivalent to averaging over plus and minus half an input frame which is equivalent to a shutter angle of 360 degrees. A shutter time of 0.5 is equivalent to a shutter angle of 180 degrees. Imagine a grey rectangle moving left to right horizontally across the


KRONOS 25Parameters

screen. Figure 11 and Figure 12 show how Shutter Time affects the retimed rectangle.

Shutter Samples - Sets the number of in-between images used to create an output image during the shutter time. Increase this value for smoother motion blur.

Automatic Shutter Time - Automatically varies the Shutter Time through-out the sequence. This is on by default.

Output - Sets the final output display for the re-timed image. Selecting anything other than Result is only useful when a Matte Component is used.

• Result - displays the retimed Source image.

• Matte - displays the retimed Matte input.

• Foreground - displays the retimed foreground - the background regions outside the matte input may show garbage.

• Background - displays the retimed background - the foreground regions inside the matte input may show garbage.

Matte Component - The Matte input can be used to help the motion estimation algorithm understand what is foreground and background in the image so that the dragging of pixels between overlapping objects can be reduced. White areas of the matte are considered to be foreground, and black areas background. Grey areas are used to attenuate between

Figure 11. Shutter Time 1. Figure 12. Shutter Time 0.5.

Figure 13. Shutter Samples 2. Figure 14. Shutter Samples 20.


KRONOS 26Parameters

foreground and background. When the matte input is filled with an appropriate clip, this parameter controls how the pixel values in the matte are used to do the masking.

• None - don’t mask.

• Src Alpha - use the alpha of the source input.

• Src Inverted Alpha - use the inverted alpha of the source input.

• Matte Luminance - use the luminance of the matte input.

• Matte Inverted Luminance - use the inverted luminance of the matte input.

• Matte Alpha - use the alpha of the matte input.

• Matte Inverted Alpha - use the inverted alpha of the matte input.


Block Size - To find where a certain pixel is located in the analysis range, the vector generation algorithm looks for a block of pixels centered around that pixel. Block size defines the width and height of these blocks (in pixels). On rare occasions, a large block size can produce data that’s lacking in detail. This is because a small feature can fit into a large block, causing the motion estimation to concentrate on the background motion and ignore the small feature. A small value, instead, can produce a noisy motion field, as there aren’t enough constraints in a small block to fit the motion accurately. In most cases, however, the default value is small enough so that details aren’t lost, and the smoothing step of the algo-rithm ensures the motion field isn’t too noisy. Therefore, this value very rarely needs editing.

Over smooth - This is a computationally intensive smoothing operation that performs a different vector-smoothing operation to normal. This generates highly smooth vector fields (as if the vectors were combed several times with a fine-toothed comb), but may also sacrifice a lot of required detail. In most cases, this level of smoothing isn’t necessary. However, if you have a problem sequence that looks too ’broken up’ in the retime, you may want to toggle Over smooth on.

Filtering - Sets the quality of filtering when producing in-between frames.

• Extreme - uses a sinc interpolation filter to give a sharper picture but takes a lot longer to render.

• Normal - use bilinear interpolation which gives good results and is a lot quicker than Extreme.

Warp Mode - Sets how to control the new timing of the clip.

• Simple - this is the quickest option, but may produce less than optimal results around moving objects and image edges.


KRONOS 27Parameters

• Normal - this is the standard option, with more optimal treatment of moving objects and image edges.

• Occlusions - this is an advanced option that can improve the results when not doing a separated picture build with multiple vector sequences and mattes. It attempts to reduce the level of background dragging that occurs between foreground and background objects.

• Sharp Occlusions - This option is similar to Occlusions, but produces fewer artifacts where the vector fields are generated from 3D CG sources.

Show Vectors - Switch this on to display the vectors on the screen.

Luminance Correct - LME is highly dependent upon the idea that the brightness of objects doesn’t vary through a sequence. Where brightness varies rapidly - for example a highlight moving across the bodywork of a car - the motion calculation will perform poorly. The luminance of a shot can come from other sources too - such as an overall flicker problem. In these cases where there is a global luminance shift, toggling this control on will allow the LME algorithm to take account of overall brightness changes between frames.

Tolerances - For efficiency, much of the LME is done on luminance only - that is, using monochrome images. The tolerances allow you to tune the weight of each colour channel when calculating the image luminance. These parameters rarely need tuning. However, you may, for example, wish to increase the red weighting Weight Red to allow the algorithm to concentrate on getting the motion of a primarily red object correct, at the cost of the rest of the items in a shot.

Weight Red

Weight Green

Weight Blue


MATCHGRADE

This chapter looks at automatic colour matching using F_MatchGrade.

Introduction It is often necessary to match the colours of one clip with those of another. When filming outside at different times of the day, you will inevitably get colour and luminance differences that will have to be corrected if the sequences are to be composited or edited together.

You can, of course, use colour correction tools and trial and error to try and match the clips. But this tends to be time-consuming and requires some considerable skill. F_MatchGrade does it all for you by automatically modifying the colour histogram of an image to match a reference image. This plug-in can also be used to boost desaturated images.

Quick Start To match the colours of one clip with those of another, do the following:

1. Select Image > Read to load the clip whose colours you want to change (source clip). Then, load the clip whose colours you want to match (refer-ence clip).

2. Select Furnace > F_MatchGrade.

3. Connect the source clip to F_MatchGrade’s Apply To input, and the reference clip to the Target input.

Figure 15. Source image. Figure 16. Reference image.

Figure 17. Output image.

MATCHGRADE 29Inputs

4. View the output which should now match the look of the Target input.

5. If the match isn’t close enough, try increasing Iterations. This increases the number of refinement passes. More iterations should produce a better match but will take longer.

By default, F_MatchGrade calculates the transform needed to match the source image frame to the target image frame, and applies this transform to every frame of the source sequence. This way, the transformation is temporally uniform. However, if you want the colour transfer to update according to the current frame of the source and target clips, you can set the Source Colour From and Target Colour From parameters to Current Frame.

If you have already spent some time on a specific frame to make it look right and want to apply the same colour transform to the rest of the clip, do the following:

1. Connect the clip whose colours you want to adjust to F_MatchGrade’s Apply To input.

2. Connect the single original frame to the source colour (Colour) input, and the modified frame to the target colour (Target) input.

F_MatchGrade will calculate the transform needed to match the Colour frame to the Target frame, and apply this transform to every frame of the Apply To sequence.

Inputs F_MatchGrade has three inputs:

• The first, Apply To input, is the sequence to which a colour transform will be applied.

• The second, Target input, is the sequence whose colours F_MatchGrade will try to match. If only the first two inputs are supplied, this transform will make each frame of the Apply To sequence match the Target input.

• If a third, Colour input, is supplied, F_MatchGrade will calculate the transform that matches the Colour input to the Target input, and then apply that transform to each frame of the Apply To sequence.


Iterations - The number of refinement passes. More iterations should produce a better match but will take longer.

Advanced - the lesser used refinement controls.


MATCHGRADE 30Parameters

Source Colour From - sets how the source clip is used to calculate the colour transformation. This parameter is only available if a Source input is connected.

• Specified Frame - the colour transformation is calculated using a single frame from the source clip. You can specify this frame using the Source Frame parameter below. This is the default.

• Current Frame - the colour transformation is calculated so that it updates according to the frames in the source clip.

Source Frame - sets the source clip frame that is used to calculate the colour transformation. By default, this is set to frame 0, which means frame 1 in NukeX. This parameter is only available if a Source input is connected.

Target Colour From - sets how the target clip is used to calculate the colour transformation.

• Specified Frame - the colour transformation is calculated using a single frame from the target clip. You can specify this frame using the Target Frame parameter below. This is the default.

• Current Frame - the colour transformation is calculated so that it updates according to the frames in the target clip.

Target Frame - sets the target clip frame that is used to calculate the colour transformation. By default, this is set to frame 0, which means frame 1 in NukeX.


MOTIONBLUR

This chapter looks at adding motion blur using F_MotionBlur.

Introduction F_MotionBlur uses the Foundry’s advanced motion estimation technology to add realistic motion blur to a sequence.

F_MotionBlur uses the same techniques and technology as the motion blur found in F_Kronos (see “Kronos” on page 21), but presents the controls in a less complex, more user friendly way. However, if you need precise control over the motion vectors used for adding blur, or a large temporal range (that is, a very high shutter time), you should use F_Kronos.

Quick Start To add motion blur to a sequence, do the following:

1. Select Image > Read to load the clip you want to add motion blur to.

2. Select Furnace > F_MotionBlur to apply F_MotionBlur. View its output.

3. Select a suitable Shutter Time, depending on the amount of blur you wish to add.

4. Render the sequence to see the motion blurred result.

Figure 18. Before F_MotionBlur. Figure 19. After F_MotionBlur.

Figure 20. A low Shutter Time. Figure 21. A high Shutter Time value.

MOTIONBLUR 32Inputs

5. If you can see that the motion blur has been created from a few discreet images, try increasing Shutter Samples. This will result in more in-between images being used to generate the motion blur and so result in a smoother blur.

If your sequence is composed of a foreground object moving over a background, the motion estimation is likely to get confused at the edge between the two. To fix this, you can do the following:

1. Add a matte of the foreground region to the Matte input.

2. Use Matte Component to select which component of the matte to use.

This will force the motion of the foreground to be calculated separately to the motion of the background and so should produce less artefacts in the motion blur.

Inputs F_MotionBlur has four inputs:

• Src is the sequence to add motion blur to.

• Matte is an optional matte of the foreground to help improve the motion blur at foreground/background boundaries.

• In addition, you can supply pre-calculated motion vectors to the vector inputs, Background Vectors (BgVec) and Foreground Vectors (FgVec). You can generate pre-calculated motion vectors using

Figure 22. Shutter Samples = 2. Figure 23. Shutter Samples = 4.

Figure 24. Shutter Samples = 8. Figure 25. Shutter Samples = 30.


MOTIONBLUR 33Parameters

F_VectorGenerator (see “VectorGenerator” on page 56) or third-party software.

If you have separate vectors for the background and foreground, you should connect them to the appropriate inputs and supply the matte that was used to generate them to the Matte input. If you have a single set of vectors, you can connect it to either of the vector inputs.


Shutter Time - Sets the equivalent shutter time of the re-timed sequence. A shutter time of 1 is equivalent to averaging over plus and minus half an input frame which is equivalent to a shutter angle of 360 degrees. A shutter time of 0.5 is equivalent to a shutter angle of 180 degrees. Imagine a grey rectangle moving left to right horizontally across the screen. Figure 26 and Figure 27 show how Shutter Time affects the retimed rectangle.

Shutter Samples - Sets the number of in-between images used to create an output image during the shutter time. Increase this value for smoother motion blur.

Vector Detail - Adjust this to vary the density of the vector field. The larger vector detail is, the greater the processing time, but the more detailed the

Figure 26. Shutter Time 1. Figure 27. Shutter Time 0.5.

Figure 28. Shutter Samples 2. Figure 29. Shutter Samples 20.


MOTIONBLUR 34Parameters

vectors should be. A value of 1.0 will generate a vector at each pixel. A value of 0.5 will generate a vector at every other pixel. For some sequences, a high vector detail near 1.0 generates too much unwanted local motion detail, and often a low value is more appropriate.

Matte Component - What to use as the (optional) foreground matte for the motion estimation.

• None - don’t use a matte.

• Src Alpha - use the alpha of the source input.

• Src Inverted Alpha - use the inverted alpha of the source input.

• Matte Luminance - use the luminance of the Matte input.

• Matte Inverted Luminance - use the inverted luminance of the Matte input.

• Matte Alpha - use the alpha of the Matte input.

• Matte Inverted Alpha - use the inverted alpha of the Matte input.


REGRAIN

This chapter looks at adding grain to sequences using F_ReGrain.

Introduction F_ReGrain is used to add grain to a sequence. It has been designed to sample an area of grain from one image and then to generate unlimited amounts of this grain with exactly the same statistics as the original. This new grain can then be applied to another image.

Figure 30 shows an enlarged and exaggerated sample of grain from Kodak 320 film stock. F_ReGrain was used to sample the original Kodak 320 stock and synthesize a plate of grain. The result is shown in Figure 31. Note that the grain characteristics closely match the original.

Similarly, Figure 32 is a sample from Kodak 500 film stock and Figure 33 shows this replicated using F_ReGrain.

Colour Space Because colour space transformations can distort the look of grain, and by default, NukeX converts your footage to linear, if you are sampling grain

Figure 30. Kodak 320. Figure 31. F_ReGrain.

Figure 32. Kodak 500. Figure 33. F_ReGrain.

REGRAIN 36Quick Start

from your own plate, you need to make sure you tell the plug-in what space the plate was in originally, so the sample isn’t distorted.

This is accomplished by setting the Grain Colour Space menu in the F_ReGrain controls to the right space. For example, if you were sampling grain from a film scan, you would want to set this to Cineon. If you had footage from a digital video camera, this would most likely be sRGB. F_ReGrain will automatically set the right colour space when using one of the pre-sampled grain clips, which are in sRGB.

F_ReGrain also works best when applying grain in the same colour space that the sampled grain originally existed. There is a menu in the Response section of the F_ReGrain controls which allows you to match or override this. It defaults to applying the grain in the same space as the sample.

Quick Start

Adding Sampled Grain To add grain to a sequence, do the following:

1. Select Image > Read to load the sequence you want to add grain to. Then, load the image you want to sample grain from.

2. Make sure you are working at full resolution and not proxy resolution. F_ReGrain will not work at proxy resolution. (See “Proxy Resolutions” on page 39.)

3. Select Furnace > F_ReGrain.

4. Connect the sequence that you want to have grain to F_ReGrain’s source (Src) input. Then, connect the sequence you want to sample grain from to the Grain input. View the output from F_ReGrain.

5. Set Grain Type to From Grain Clip.

6. Position the on-screen sample region over an area of the Grain sequence just containing grain and no picture detail. See Figure 34. It is important to get your selection right. You should avoid any image detail or even a plain area that has luminance variations underneath the grain. The better this initial selection, the better the result will be. If you can’t find a decent sample area on the current frame, then try other frames from the same film stock. The default size of the sample area should be enough to gather information about the grain characteristics of your image.


REGRAIN 37Quick Start

However, you may need to change its size and shape to fit over a plain area free of image detail.

Warning! There is a minimum size of this sample area below which the statistical analysis of the grain will be unreliable. If the sample area you select is too small, you will see a warning message which prompts you to select a larger region. (See “Proxy Resolutions” on page 39.)

7. View the output of F_ReGrain to judge the results. The output will now contain the Src image with grain from the Grain image applied. Both the size and the luminance of the new grain can be manually tweaked using Grain Size and Grain Amount respectively. It helps to view the Grain input while editing the parameters of F_ReGrain.

The grain is sampled on a single frame which is set when you adjust the sample area (or by manual adjustment of the Analysis Frame parameter). Although it is sampled on only one frame, the algorithmically created grain will change from frame to frame but mirror the characteristics of the sample grain.

Using Pre-Sampled, Standard Grain Types

If you don’t have an image to sample grain from, you can also select from a

variety of pre-sampled, standard grain types. Do the following:

1. Select Image > Read to load the sequence you want to add grain to.

2. Make sure you are working at full resolution and not proxy resolution. F_ReGrain will not work at proxy resolution. (See “Proxy Resolutions” on page 39.)

Figure 34. This shows two possible selection regions that contain no edge detail and little luminance variation.


REGRAIN 38Response

3. Select Furnace > F_ReGrain.

4. Connect the sequence that you want to have grain to F_ReGrain’s source (Src) input.

5. Set Grain Type to Preset Stock.

6. Try the different grain types using the Preset Stock menu. 2K, 4K, aperture corrected, and non aperture corrected stocks are included. Individual colour channels can be selected and adjusted using the Advanced parameters.

Response In its default setting, F_ReGrain adds the same amount of grain over the whole image. However, the amount of grain on an image is normally a function of luminance. Various parameters in the Grain Response group allow you to adjust how the amount of grain added varies with luminance:

• Pressing Sample Grain Response will cause the variation of the amount of grain with luminance to be calculated from the Grain input, and switching on Use Sampled Response will apply these curves to the grain added to the Src sequence.

• To view the sampled response curves, switch on Draw Response; an example is shown in Figure 35.

• The amount of grain added to the lowlights, midtones and highlights of the image can be adjusted using the Low Gain, Mid Gain and High Gain parameters. The effect of adjusting these can also be seen on the response curves.

Figure 35. This shows an example of the grain reponse with luminance. The x axis represents luminance and the y axis

the amount of grain.


REGRAIN 39Checking the Result

Checking the Result

To test that the new grain is the same as the old grain, set Output to Grain Plate.

This generates a sheet of grain with the same luminance level as the mean of the sample region. The sample region with the original grain is also displayed. It should be impossible to differentiate between the two regions. Figure 36 shows a good selection area giving a good test plate of grain in Figure 37.

Figure 38 shows a poor selection area since it contains image detail. Figure 39 shows the resulting test plate which clearly highlights the problem.

Proxy Resolutions Grain manipulation at proxy resolution should be avoided as the results are unreliable. The grain selection area may be too small at proxy resolution to give a good result, and making this area larger may drag in unwanted detail from the image. If you try to use F_ReGrain at proxy resolution, we simply pass the image through untouched and issue the following warning:

Cannot work at proxy scale.

Figure 36. Good selection area... Figure 37. ...producing a good test plate of grain, free of

artefacts.

Figure 38. Bad selection area... Figure 39. ...producing a poor result.


REGRAIN 40Inputs

We decided that this was preferable behaviour to doing poor grain replication at proxy resolution. You can, of course, crop the input clip and work with that rather than the proxy. There is a minimum size for the selection box, which is about 37x37 at base resolution. If the box you select is smaller, you will get this warning along the top of the viewer:

Sample box is too small - please select a larger sample of the grain.

Inputs F_ReGrain has two inputs:

• The source (Src) is the image to which the grain will be added.

• The Grain is the image from which grain will be sampled. When a Grain input is supplied, the plug-in will automatically switch to using grain sampled from this input. However, the supplied grain stocks are still available.

Parameters The parameters for this plug-in are described below:

Grain Type - Selects whether the grain is sampled from the Grain image (From Grain Clip) or from a set of standard stocks.

• Preset Stock - grain characteristics are sampled from a supplied film stock. 2K, 4K, aperture corrected and non aperture corrected stocks are supplied. Although standard stocks are included, it is recommended where possible that you sample from the film stock you are trying to match.

• From Grain Clip - samples and reconstructs the grain characteristics from the Grain input.

Preset Stock - The film stock the grain characteristics are sampled from when Grain Type has been set to Preset Stock. Common Fuji and Kodak stocks are supplied. The exposure can be under, over, or, if left blank, non aperture corrected. The size is either 2K or 4K pixels. For example, FUJIF500 2K refers to the grain characteristics sampled from a 2K plate of Fuji Film 500 film stock non aperture corrected.

Grain Amount - Adjusts the brightness of the grain. Setting this to 0 means no grain is added.

Grain Size - Adjusts the size of the grain granules. The larger the value, the bigger and softer the granules.

Output - Sets whether to render the result or a test image.


REGRAIN 41Parameters

• Result - shows the Source image with the grain applied.

• Grain Plate - shows a test image with the grain applied. This test image is composed from a section of the input image surrounded by a uniform solid colour sampled from the image with the grain applied (Figure 39 on page 39). If the inner area is indistinguishable from the outer area, then you have a good grain sample (Figure 37 on page 39).

Analyse - This is a push button which will trigger an analysis of the input clip. Press this button if the input clip from which the grain was analysed has changed but you do not want to move the analysis region to trigger re-analysis. Whenever the input clip changes, you will see the following warning in the Viewer:The clip from which the grain was analysed has changed. Press Analyse or move the analysis region to re-analyse grain.

Analysis Region - A selection box that marks the region of image used to analyse the grain when Grain Type is set to From Grain Clip. This part of the frame must contain no image detail, only grain. (Figure 36 on page 39.)



Analysis Frame - sets the frame to sample the grain from.

Grain Colour Space - This tells F_ReGrain what colour space the grain sample clip was in when the grain originated. Setting this correctly ensures that the grain is not exaggerated by any colour space conver-sions prior to sampling.

• Cineon

• sRGB

• Linear

Advanced - The parameters under Advanced allow detailed adjustment of the grain.

Process Red - Switch this on to process the red channel.

Red Amount - sets the brightness of the grain in the red channel.

Red Size - adjusts the size of the grain granules in the red channel.

Process Green - Switch this on to process the green channel.

Green Amount - sets the brightness of the grain in the green channel.

Green Size - adjusts the size of the grain granules in the green channel.

Process Blue - Switch this on to process the blue channel.

Blue Amount - sets the brightness of the grain in the blue channel.

Blue Size - adjusts the size of the grain granules in the blue channel.


REGRAIN 42Parameters

Grain Response - The parameters under Grain Response allow the amount of grain added to be varied as a function of the image luminance.

Apply Grain In - This controls what colour space the grain sample is re-applied to the image.

Generally, this should be set to Grain Colour Space to ensure the most accurate recreation.

You may want to override this though for some looks or special cases.

• Cineon / sRGB / Linear - The grain sample will be applied in the specified space.

• Grain Colour Space - The Grain sample will be applied in the colour space set in the Grain Colour Space menu, in the Grain Sample section.

Low Gain - adjusts the gain of the grain in the lowlights.

Mid Gain - adjusts the gain of the grain in the midtones.

High Gain - adjusts the gain of the grain in the highlights.

Use Sampled Response - switch this on to scale the brightness of the grain as a function of the luminance of the Grain image.

Sampled Response Mix - this control is usually set to 1. Decreasing it reduces the effect of the response curves until, at 0, they have no effect on the output. This parameter is only available if Use Sampled Response is on.

Sample Grain Response - press this to update the response curves from the current frame. Multiple presses accumulate the grain response rather than resetting every time. This parameter is only available if Use Sampled Response is on.

Reset Grain Response - press this to reset the grain curves to their default (flat) response. This parameter is only available if Use Sampled Response is on.

Draw Response - overlays the response curves on the bottom left corner of the viewer. This parameter is only available if Use Sampled Response is on.


RIGREMOVAL

This chapter looks at the removal of unwanted objects (rigs) from image sequences without accurate rotoscoping or keying to produce a clean plate.

Introduction In this context, we define a rig as a foreground element in a sequence that moves over a background element. The plug-in will only work satisfactorily if it is possible to model the background motion by a global 3D transform. For example, if the background contains multiple objects moving in different directions, the results will be poor. Typically, good results will only be achieved in situations where a skilled artist could generate, and track in, a clean plate in order to repair the sequence. However, this plug-in should make the process quicker and easier. The rig removal algorithm works by estimating the background motion between successive frames, ignoring the foreground object, and then using the motion information to look forward and backward in the sequence in order to find the correct piece of background to fill in the missing region. The size of the missing region and the speed of the background dictate how far away from the current frame it is necessary to search to find the correct information.

Figure 40. Before with bird. Figure 41. After applying F_RigRemoval.

Figure 42. Before with taxi. Figure 43. After applying F_RigRemoval.

RIGREMOVAL 44Quick Start

Quick Start To remove an unwanted object from an image sequence, do the following:

1. Select Image > Read to load the sequence with an unwanted object.

2. Select Furnace > F_RigRemoval and connect your image sequence to F_RigRemoval’s Src input. View the output of F_RigRemoval.

3. On each frame, define the area that will be repaired. You can do this in the following three ways:

• If the image sequence has an embedded alpha channel, then you can use that alpha to define the area to be repaired. To do so, set Rig Region to Src Alpha.

• Using F_RigRemoval’s RigMask input, you can feed in a matte sequence to define the area to be repaired. To use this matte, set Rig Region to Rig Mask Alpha (or one of the other Rig Mask options).

• You can use the on-screen rectangle to define the area to be repaired. To do so, set Rig Region to Box. Position the on-screen rectangle on top of the unwanted object. To key-frame the position to follow the object, select Set key from the animation menus next to Rig Region BL and Rig Region TR in the Rig Region Box parameter group. Move to a new frame and reposition the on-screen rectangle. A new key frame is set automatically. Repeat as necessary until the rectangle covers the object on every frame you want to remove the object from.

Whichever method you choose, the region does not need to be the exact foreground region, but just a rough outline. However, you should avoid making it unnecessarily large as this will increase rendering time.

4. Having defined the region to repair throughout the clip, set Frame Range to the number of frames that the plug-in needs to analyse forwards and backwards to find enough data to repair the sequence. On the first frame, this will be quite time consuming as the algorithm needs to estimate the motion between each pair of frames. Subsequent frames will be much quicker.

5. If it has not been possible to replace all the foreground pixels, either because Frame Range was set too low or the background information does not exist anywhere within the sequence, the pixels will be displayed in red. Try to adjust Frame Range until no red pixels are visible and then render the sequence.

In Figure 44, we are using a box to define the pixels to replace, and Frame Range is set to zero. Increasing this value, as shown in Figure 45,



gathers pixels from other frames and improves the result. To completely remove the red pixels, you’d need a Frame Range value of 5.

6. View the results.

Tip F_RigRemoval is fairly slow to process, which can make the initial keyframing of the rectangular region frustrating.

Sometimes, the easiest way to adjust the region is to load up F_RigRemoval, and to view the source so the effect is not processed, but the parameters are visible. Then, animate the rectangular region over the foreground object you’re trying to remove throughout the sequence. When you’re happy with the region position, click back onto F_RigRemoval’s output and wait for it to update. Slowly increase the Frame Range parameter until the whole region is repaired, then check the output on other frames.

Note F_RigRemoval uses frames from a wide range around the current frame. Feeding the output of one F_RigRemoval node into another will greatly increase the memory over head, as the second F_RigRemoval node will require the first F_RigRemoval node to calculate all its frames before passing the result on. Therefore, it is strongly recommended that you

Figure 44. Frame Range = 0. Figure 45. Frame Range = 3.

Figure 46. Original image. Figure 47. The output of F_RigRemoval.



process out the first result and then use the processed result with the second F_RigRemoval node.

Occlusions The algorithm used in F_RigRemoval is unable to differentiate between multiple foreground objects. If there is another foreground object in the sequence that moves through the background region that is being used in the repair, this second foreground object will also be cut up and used, resulting in an incorrect repair. To try and assist in these situations, it is possible to mark regions of the image as not to be used for repair by setting their alpha value to mid grey. This will ensure that spurious bits of other foreground objects do not appear in the repair.

In Figure 48, we are trying to remove the woman in the centre of the screen as she walks from left to right down the street. At this frame, a man walks in the opposite direction and her feet and his head overlap.

Figure 49 shows the normal matte for the woman, and Figure 50 shows the result of using this in F_RigRemoval. Note that the man’s head interferes with the repair and the reconstruction of the pavement is particularly bad, probably due to the man becoming the dominant motion source.

To fix this, we can adapt the matte to include a mid grey area over the man.

Figure 48. Original shot.

Figure 49. The normal matte. Figure 50. The output.


RIGREMOVAL 47Inputs

This tells the rig removal algorithm to ignore that area in the repair. This matte is shown in Figure 51, and the result is shown in Figure 52. Note that the repair on the pavement is improved, and the man is simply clipped rather than being used in the repair.

Inputs F_RigRemoval has two inputs:

• The source (Src) image with the unwanted object (rig). The source may contain an alpha channel to define the rig area.

• The optional RigMask to designate the rig area to be removed.


Rig Region - Defines the area to repair.

• Box - repair the area inside a rectangular box, controlled by the box parameters below or the on-screen box.

• Src Alpha - repair the region defined by the alpha of the source input.

• Src Inverted Alpha - repair the region defined by the inverted alpha of the source input.

• RigMask Luminance - repair the region defined by the luminance of the Rig Mask input.

• RigMask Inverted Luminance - repair the region defined by the inverted luminance of the Rig Mask input.

• RigMask Alpha - repair the region defined by the alpha of the Rig Mask input.

• RigMask Inverted Alpha - repair the region defined by the inverted alpha of the Rig Mask input.

Frames Searched - Sets whether to search forwards, backwards, or in both directions to find missing data.

Figure 51. The matte with a mid-grey area.

Figure 52. The output.


RIGREMOVAL 48Parameters

• Forward and Backward - searches before and after the current frame.

• Forward - searches frames after the current frame.

• Backward - searches frames before the current frame.

Frame Range - Sets the number of frames the algorithm should look forwards and backwards in the sequence to find the missing data. If you are getting red pixels, then increase this value. See Figure 45 on page 45.

Frames Used in Range - If Frame Range has to be set to a large number to make an effective repair, the rendering time can be prohibitive. Frames Used in Range can speed up the repair by not using every frame to fill the foreground region, effectively skipping frames. However, this may reduce the quality of the result.

• All Frames - use every frame in the searched frame range to construct the repair.

• Half of Frames - use every other frame in the searched frame range to construct the repair.

• Quarter of Frames - use every fourth frame in the searched frame range to construct the repair.

• 10% of Frames - use every tenth frame in the searched frame range to construct the repair.

• Max 25 Frames - use no more than 25 frames from the searched frame range to construct the repair. This option can be useful if Frame Range has been set to a very large number.

Max Rig Movement - To avoid perspective changes, F_RigRemoval searches for the missing data inside an area immediately around the rig region. Max Rig Movement defines the width of this area (in pixels). Fast movement in the Src footage requires a higher value than slow movement. However, because the area used for the repair may be from a different part of the image, high values can introduce perspective problems.

Rig Region Box - The rectangular area used to define the repair when Rig Region is set to Box.

Rig Region BL - controls the position of the bottom left corner of the rig region.

Rig Region TR - controls the position of the top right corner of the rig region.





RIGREMOVAL 49Parameters



Luminance Correct - Switch this on to correct for luminance changes from information taken from other frames. This is particularly important if the lighting changes throughout the sequence.

Perspective Correct - Switch this on to correct for minor perspective changes.

Overlap Correct - The repair is built up using slices of information from other frames in the sequence. These slices can be overlapped and blended to give a more natural looking repair. This parameter controls how much the regions overlap. Increasing this parameter too much will degrade image sharpness.

Repair Fail Marker Opacity - Sets the level of transparency of the red pixels used to show where the repair has failed.

Preserve Alpha - Switch this on to preserve the original alpha channel. By default, this is switched off, and the alpha channel is set to white where the repair has failed and black everywhere else.


STEADINESS

This chapter looks at how to stabilise a shot using F_Steadiness.

Introduction F_Steadiness uses Global Motion Estimation (GME) to calculate a four-corner pin, so that camera motion within a single shot can be smoothed out over a range of frames or removed by locking to a specific frame.

F_Steadiness needs to analyse the input clip before it can render useful output. This analysis is done when you press Analyse in the F_Steadiness controls. During the analysis, F_Steadiness key-frames a four-corner pin which will stabilise the clip in subsequent renders. Without having performed an analysis pass, F_Steadiness will not do anything useful on render.

F_Steadiness can work in two ways. These are:

1. Smooth - A range of frames around each frame is analysed for motion and an average of that motion used to calculate the corner pin. Use this to keep the overall camera motion, but to smooth out sharp bumps and kicks.

2. Lock - A lock frame is specified and steadiness attempts to register each individual frame to that lock frame. Use this to completely remove camera motion from the sequence.

In lock mode, each frame in the clip must share a substantial amount of the scene with the lock frame, so you can’t lock each frame in a 360 degree pan to the first one. However, in smooth mode, because F_Steadiness is only working on a small window of frames around the current frame, you can use it in shots which change completely over time.

The analysis region is used to control which section of the reference frame is being matched to each source frame. In lock mode, the reference is the lock frame, so you should position the analysis region when looking at the lock frame. In smooth mode, it looks at the incremental differences between frames, in which case you should place the analysis region in the area you want to appear with smooth motion.

The transformation in F_Steadiness is concatenated with other NukeX transform nodes. This means that if you add a row of F_Steadiness and NukeX transform nodes to a tree, their functions are combined. Because the image is only resampled once, there is no loss of image quality and

STEADINESS 51Quick Start

processing time is decreased. However, as in NukeX, certain nodes, including colour correction nodes, can break the concatenation.

For an overview of Global Motion Estimation, and a description of the common way Global Motion Effects work, please see “Global Motion Estimation” on page 71.

Quick Start This section gives a very brief outline of how to use the plug-in.

Smoothing Out Camera Motion

To keep the overall camera motion but to smooth out sharp bumps and kicks, do the following:

1. Find a shot that has some camera shake in it and select Image > Read to load it.

2. Select Furnace > F_Steadiness to apply F_Steadiness, and view its output.


No valid key framed analysis found, please press Analyse.

3. In the Advanced F_Steadiness controls, make sure Mode has been set to Smooth.

4. Click on the Analyse button.

F_Steadiness will now start analysing each frame in the shot, figuring out the smoothing corner pin and writing it as key frames to the corner pin parameters, Bottom Left, Bottom Right, Top Left, and Top Right.

After a brief pause (while F_Steadiness calculates the transforms in the frame range), F_Steadiness will update the timeline and you will see the steadied image render in the viewer.

If at any point you interrupt the analysis, the pins it has calculated until that point will be retained.

5. Play or scrub through the stabilised frames.

6. If you want to make the result smoother, increase the Smoothing parameter in the Advanced parameter group and the corner pin will be recalculated immediately to give a smoother shot. You don’t need to re-analyse the sequence if you do this; as F_Steadiness has kept the raw inter-frame transforms cached away, all it needs to do is re-write the keys for the average smoothing pin.

Locking To A Frame To completely remove camera motion from a sequence, do the following:


STEADINESS 52Inputs

1. Find a shot that has camera shake but where all frames share scene information and select Image > Read to load it.

2. Select Furnace > F_Steadiness to apply F_Steadiness, and view its output.


No valid key framed analysis found, please press Analyse.

3. In the Advanced F_Steadiness controls, set Mode to Incremental Lock.

4. Choose a frame somewhere in the middle of the sequence that you want to lock to, and set the Lock Frame parameter to that frame.

5. Scrub back and forth through the shot and look for a region of the shot that is shared by all frames and doesn’t change much (for example, avoid a region with people walking in front of it).

6. Whilst looking at the lock frame, position the on-screen widget for the Analysis Region over that region.

7. Hit Analyse.

The effect will start to analyse, working first forward from the lock frame, then backwards from it, until all frames to be analysed are done.

The timeline will immediately update to give you feedback in the viewer.

Inputs F_Steadiness has a single input: the shot to stabilise.

Parameters The parameters for this plug-in are dewscribed below.

Analyse - This is a push button which will trigger an analysis of the input clip and calculate a corner pin. Interrupting the analysis will not delete the corner pin keys that have already been calculated.

Render During Analysis - If set, this toggle will cause the effect to update the time line and render a freshly analysed frame so you can see the progress of the effect. Doing so will slow down the analysis somewhat, so toggle this off to speed up the general analysis.

Clear Analysis - Pressing this push button will delete all key frames from the corner pin parameters, allowing you to force a re-analysis if you feel the need to.



STEADINESS 53Parameters













Mode - This parameter controls whether F_Steadiness is smoothing the shot while keeping the overall camera motion, or locking the shot to a single frame to completely remove camera motion. It can be set to:

• Incremental Lock - in which case, it will calculate the pin that takes each frame to the lock frame. This calculates the pin by working from the lock frame out to each frame, calculating the GME between each frame incrementally and accumulating it to create the corner pin.

• Absolute Lock - this also calculates a pin that takes each frame to the lock frame. However, it does so by doing GME directly from the frame in question directly to the lock frame.



• Smooth - in which case, we are smoothing the shot for a range of frames described by the Smoothing parameter. This is the default value. It can be used to keep the overall camera motion, but to smooth out sharp bumps and kicks.

Incremental locks work better in some situations, whilst absolute locks work better in others. However, absolute locks, when they work, are often more accurate. If, for example, you want to get a lock between frame 0 and frame 100, try using absolute lock first. If absolute lock doesn't work, it can be because frame 0 and frame 100 are too different. In this case, incremental lock can help if there is a gradual spatial progression from frame 0 to frame 100.

Smoothing - This controls the range of frames to average motion over when Mode is set to Smooth. The default value is 10.

Lock Frame - This controls the frame which will be locked to when Mode is set to either of the lock modes. By default, this is set to 0. Note that what is frame 0 for F_Steadiness is frame 1 in NukeX. Therefore, if you are looking at frame 3 in the Viewer and want to use that frame as the lock frame, you need to enter 2 as the Lock Frame value.

Accuracy - This controls the time/accuracy trade off. The higher this is, the slower the analysis, but you have a better likelihood of a good result.





Invert - If set, then the inverse of the calculated four-corner pin is used during render. This works best with the lock modes, and can be used to track static locked-off plates back into a shot.

Auto Scale - To smooth out or remove camera motion, F_Steadiness translates and rotates the frames in the source clip. This leaves black pixels around the image. The Auto Scale parameter lets you fill in the black gaps at the edges by scaling the output image up. A value of 1 (the default) scales the image up until no black is visible, whereas a value of 0 disables scaling and leaves the black edges untouched. Auto Scale uses the minimum scale necessary to remove the black gaps at the edges to preserve as much detail as possible.




Bottom Left - the lower left corner pin.

Bottom Right - the lower right corner pin.

Top Left - the upper left corner pin.

Top Right - the upper right corner pin.


VECTORGENERATOR

This chapter looks at producing images containing motion vector fields using F_VectorGenerator. For more information about motion vector fields, please refer to the chapter on Local Motion Estimation on page 76.

Introduction Many of the FurnaceCore plug-ins rely on motion estimation to produce their output. This means there is some overlap in the work they do: if you have more than one FurnaceCore effect in your project with one or more of the same inputs, they might well be performing identical motion estimation tasks. F_VectorGenerator is a utility plug-in designed to save processing time by allowing you to perform the motion estimation separately, so that the results can then be reused by other FurnaceCore effects.

In general, once you have generated a sequence of motion vector fields that describe the motion in a particular clip well, they will be suitable for use in any FurnaceCore plug-ins which can take vector inputs. These include F_Kronos and F_MotionBlur.

We recommend that you insert a Write node after F_VectorGenerator to render the original images and the vector channels as an .exr file. This format allows for the storage of an image with multiple channel sets embedded in it. Later, whenever you use the same image sequence, the motion vector fields will be loaded into NukeX together with the sequence and are readily available for the FurnaceCore plug-ins. For information on how to generate motion vector fields and render them as an .exr file, see “Quick Start” on page 57.

Output The output from F_VectorGenerator consists of two sets of motion vectors for each frame. These are:

1. The backward vector field: the x and y offsets per pixel that, when applied to the previous frame in the sequence, allow you to reconstruct an approximation to the current frame.

2. The forward vector field: the x and y offsets needed to transform the next frame into an approximation to the current one.

The closeness of the approximation in each case can be controlled by altering the plug-in’s parameters.

VECTORGENERATOR 57Quick Start

Note that the output from F_VectorGenerator will be stored in the vector channels, so you might not see any image data appear when you first render the effect. To see the output inside NukeX, you need to select motion, forward, or backward from the channel menu at the top left of the viewer (or press page down until the data appears). Figure 54, Figure 55, and Figure 56 show the different vector images for the jet clip shown in Figure 53.

Quick Start To generate motion vector fields for an image sequence, do the following:

1. Select Image > Read to load the sequence you wish to calculate the motion in.

2. Select Furnace > F_VectorGenerator. Connect your image sequence to F_VectorGenerator’s Source input.

F_VectorGenerator calculates the motion vector fields and stores them in the vector channels. If you want to view the motion vector fields, connect a Viewer to F_VectorGenerator’s output and select foreground or back-ground from the channels menu.

3. Select Image > Write to render the motion vector fields. If you want to embed the vector channels in the image sequence, set channels to all and file type to exr in the Write node controls. This way, the newly created vector channels will be saved in the channels of the image sequence. When you need to manipulate the same sequence again later, the motion vectors will be loaded into NukeX together with the sequence.

Figure 53. Jet sequence. Figure 54. Forward and backward motion vectors for the jet

sequence.

Figure 55. Forward motion vectors.

Figure 56. Backward motion vectors.


VECTORGENERATOR 58Inputs

4. If the calculated motion vector fields do not produce the results you’re after when used with the other FurnaceCore plug-ins, use the F_VectorGenerator parameters to adjust the way the vector fields are calculated. The available parameters are described below.

Inputs F_VectorGenerator has two inputs:

• The sequence from which to generate motion vectors, Source.

• An optional foreground matte (Matte). This can be used to help the motion estimation algorithm inside F_VectorGenerator understand what is foreground and background in the image, so that the dragging of pixels between overlapping objects can be reduced. White areas of the matte are considered to be foreground, and black areas background. Grey areas are used to attenuate between foreground and background.

Parameters The parameters for this plug-in are described below. With the exception of Output, all of the parameters control the Local Motion Estimation algorithm used inside F_VectorGenerator, so please see the chapter on Local Motion Estimation on page 76 for a more detailed description of their effects.

Vector Detail - This determines the accuracy of the motion vectors. The maximum value of 1 will generate one vector per pixel. This will produce the most accurate vectors but will take longer to render.

Smoothness - Vector fields usually have two important qualities: they should accurately match similar pixels in one image to another and they should be smooth rather than noisy. Often, it is necessary to trade one of these qualities off against the other. A high smoothness will mean the output vector fields are less detailed, but is less likely to provide you with the odd spurious vector (think of it as if the vectors were combed until they are smooth with respect to their neighbours). A low smoothness will concentrate on detail matching, even if the resulting field is jagged. The default value of 0.5 should work well for most sequences.

Output - When a matte input is supplied, this determines whether the motion vectors corresponding to the background or the foreground regions are output.

• Foreground - the vectors for the foreground motion are output.

• Background - the vectors for the background motion are output.

Matte Component - where to get the (optional) foreground mask to use for motion estimation.


VECTORGENERATOR 59Parameters

• None - don’t use a foreground mask.

• Source Alpha - use the alpha of the source.

• Source Inverted Alpha - use the inverted alpha of the source.

• Matte Luminance - use the luminance of the matte.

• Matte Inverted Luminance - use the inverted luminance of the matte.

• Matte Alpha - use the alpha of the matte.

• Matte Inverted Alpha - use the inverted alpha of the matte.


Block Size - To find where a certain pixel is located in the analysis range, the vector generation algorithm looks for a block of pixels centered around that pixel. Block size defines the width and height of these blocks (in pixels). On rare occasions, a large block size can produce data that’s lacking in detail. This is because a small feature can fit into a large block, causing the motion estimation to concentrate on the background motion and ignore the small feature. A small value, instead, can produce a noisy motion field, as there aren’t enough constraints in a small block to fit the motion accurately. In most cases, however, the default value is small enough so that details aren’t lost, and the smoothing step of the algo-rithm ensures the motion field isn’t too noisy. Therefore, this value very rarely needs editing.

Over Smooth - This is a computationally intensive smoothing operation that performs a different vector-smoothing operation to normal. This generates highly smooth vector fields (as if the vectors were combed several times with a fine-toothed comb), but may also sacrifice a lot of required detail. In most cases, this level of smoothing isn’t necessary.

Tolerances - By default we analyse motion based on the brightness of the image. Specifically this is the mean of the red, green and blue channels. However, we can bias this using the red weight, green weight, and blue weight parameters. For example, if we set the red and green weight parameters to zero, we will only look for motion in the blue channel.

Weight Red - sets the contribution the red channel will make in the cal-culation of the motion vectors.

Weight Green - sets the contribution the green channel will make in the calculation of the motion vectors.

Weight Blue - sets the contribution the blue channel will make in the cal-culation of the motion vectors.


WIREREMOVAL

This chapter looks at the removal of wires from images using F_WireRemoval.

Introduction Many effects movies feature complicated stunts that require an actor to be suspended by wires for their safety. These wires need to be digitally removed.

There are many ways of doing this, including painting the wires out frame by frame and replacing large parts of the background to cover the wires. The method used depends on the type of image under the wire. F_WireRemoval is particularly good at replacing the painting method but it also includes tools to assist in clean plating when new backgrounds have to be tracked into place.

Background

Clean Plates The use of clean plates in wire removal is very common and gives good results in certain situations.

Consider a scene shot with an actor suspended from wires and then the same scene shot again without the actor. This second sequence is called the clean plate. The wires from the first shot can be painted out using pixels from the clean plate leaving the actor suspended in thin air.

Shooting a clean plate if the camera is locked off is easy. If the camera moves, then motion control rigs can be used to exactly reproduce the first pass. But it doesn’t always work, particularly if the scene is shot against

Figure 57. Three wires. Figure 58. Two of the wires have been removed using

F_WireRemoval.

WIREREMOVAL 61Background

backgrounds that don’t look the same on the second pass, such as clouds, sky, or smoke. Motion control rigs are also expensive and that makes them a rarity. Often, a clean plate is not shot during the filming and the compositor is left to create one by merging together unobstructed pixels from many frames. This single frame can then be tracked into place to cover the wires.

FurnaceCore FurnaceCore’s wire removal plug-in should make the process of wire removal much easier. It is particularly good at removing wires over heavily motion blurred backgrounds or wires over smoke, dust, or clouds. It can be used to remove each wire in a sequence or to quickly create a clean plate which can then be tracked into place.

The reconstruction of the background behind the wire can be done spatially, in other words, using only information from the current frame. Alternatively, motion estimation techniques can be used to warp frames from before and after onto the current frame so that, where available, background pixels from these frames can be used to improve the repair. Our reconstruction methods are unique in that they remove the wire without removing and reapplying the grain. They are also tuned to remove long thin objects, leaving other objects untouched. For example, if transient foreground objects cover the wire, they will be left alone.

Reconstruction Methods

Our four reconstruction methods are:

• Spatial

• Temporal With Static Scene

• Temporal With Moving Scene

• Clean Plate.

The spatial method takes the background information from adjacent pixels in the current frame, and the clean plate method takes the information from a separate clean plate input. The other methods are temporal and attempt to get background information from frames either side of the current frame. Where this information is not available, for example, because the wire covers part of the same region in one or more of the other frames, the spatial method will be used for the repair.

The spatial method is fastest. It uses a slope-dependent filter that interpolates across the wire at the most likely angle, given the image behind the wire. It works well when the wire is over backgrounds such as sky, clouds, or smoke, and can also cope with some backgrounds where there is


WIREREMOVAL 62Background

a noticeable gradient, such as the edge of a roof, building, or car. If this fails and the wire is moving relative to the background, you should try one of the temporal methods. These look at frames before and after the current one to find likely pixels with which to repair the region and use the spatial method in areas where there are none available.

Where traditional clean plates are possible or give better results than using F_WireRemoval to repair the wire on each frame, you can supply a clean plate as the fourth input to the plug-in. It will then automatically matchmove it to repair each frame. If the overall luminance of the background is different to that of the clean plate or varies during the sequence, turn on Luminance Correct. The luminance of the background pixels used for the reconstruction will then be adjusted before the repair is made. Of course, various FurnaceCore tools can be used to create a clean plate, including an F_WireRemoval pass on a single frame where the repair has been successful.

Tracker F_WireRemoval incorporates a tracker which can automatically track a moving wire through a clip. This tracker has its own control panel, which will float inside the NukeX viewer if you have checked Show On Screen Controls in the WireRemoval controls. Figure 59 is a screenshot which illustrates this.

You can resize the panel by dragging its edges. For details of the rest of the controls, see the section on Tracker Controls on page 65. All the controls are also available in the WireRemoval properties.

Figure 59. Screenshot with tracker panel.


WIREREMOVAL 63Quick Start

Quick Start To remove a wire from an image sequence, do the following:

1. Select Image > Read to load the clip that needs a wire removed.

2. Select Furnace > F_WireRemoval. Connect your image sequence to F_WireRemoval’s Source input. View the output from F_WireRemoval.

3. Make sure Output is set to Source, and use the on-screen tools to draw a region that defines the position and shape of the wire to be removed. For more information, see “Positioning the On-Screen Wire Tool” below.

4. If the wire you want to remove is moving, start the wire tracker by clicking on the Track Forwards button, so that it follows the wire during the clip. For more information, see “Tracking” on page 64.

5. From the Repair menu, choose a repair method that removes the wire. If you choose to use the Clean Plate method, connect a clean plate image to F_WireRemoval’s CleanPlate input. For more information on the available methods, see “Reconstruction Methods” on page 61.

6. To see the output, set Output to Repair.

7. If necessary, increase Overall Width to expand the width of the repair region until the wire disappears.

Positioning the On-Screen Wire Tool

To position the on-screen wire tool, do the following:

1. In the F_WireRemoval controls, set Output to Source. This way, you will be able to see the wire you’re trying to remove but won’t have to wait for F_WireRemoval to repair the wire every time you change wire tool’s posi-tion.

2. Choose the number of points that are needed to describe the wire: two for straight lines, three for simple curves, or five for more complex shapes. This can be done by changing the Type parameter in the Points parameter group, or by toggling the Number of Points button in the on-screen tracker panel. To show the on-screen tracker panel, check Show On Screen Controls in the WireRemoval properties.

3. Position the on-screen wire tool so that it roughly fits the wire you want to remove. To move the entire tool, drag its central line to a new location. To move an individual wire point on the tool, drag its centre. To move an individual point in either a horizontal or vertical direction only, drag the horizontal or vertical line on the cross that marks the point.

4. Press the Snap To button in the WireRemoval properties or the on-screen tracker panel. This will find the edges of the wire and adjust the removal region to fit it more closely. It does this by locating the areas of highest gradient, which should correspond to the edges of the wire.


WIREREMOVAL 64Tracking

5. In practice, the edges of the wire will be slightly soft, so the resulting region might not cover the whole width of the wire. To correct for this, you can adjust the Overall Width parameter. This expands the region outwards, over all key frames, to ensure the entire width of the wire is covered.

Tracking F_WireRemoval incorporates a tracker that will track the region to be repaired through the image sequence.

To use the tracker, do the following:

1. Drag the on-screen wire tool to position the repair region on top of the wire on one frame. This sets a user key frame for the current frame, enabling tracking.

2. To start the tracker, press the Track Forwards button.

If you haven’t set a user key frame (step 1), the tracker cannot be started and F-WireRemoval will display the following message:

Please select a user key frame before tracking.

3. When the track is ready, check that the wire has been correctly tracked across the other frames and adjust the track if necessary. The tracking controls in the tracker panel are outlined under “Tracker Controls” on page 65. Any points initially positioned off the image will remain off the image. This way, the tracker will not lose the wire at the edges of the image.

4. If after tracking the wire you adjust the position of the on-screen wire tool manually on one frame, you may want to create a smooth transition between the user key frame you created and the track key frames around it. The easiest way to do this is to go on the previous and next frame on the timeline and press the Delete track key frame button. This deletes the track key frames around the frame you adjusted and forces NukeX to interpolate smoothly around the adjustment.

Similarly, if there is too much jitter in the track between some frames, you can delete the track key frames between these frames. This way, NukeX will interpolate smoothly from one frame to another.

If it is not possible to track the wire automatically, the wire can be manually key-framed. Adjusting the on-screen wire tool on each frame will automatically set user key frames. Alternatively, they can be set by pressing the Create User Key frame button in the tracker panel.


WIREREMOVAL 65Inputs

User Key Frames and Track Key Frames

User key frames are key frames you set manually. When a user key frame has been set on the current frame, this is indicated by a red key that appears on the on-screen wire tool. Track key frames are key frames set by F_WireRemoval’s tracker.

User key frames override track key frames on the same frame. This has the following consequences:

• Once you set a user key frame, the track key frame on that frame will be lost.

• When you track a wire, existing user key frames will not be replaced by track key frames.

Inputs F_WireRemoval has two inputs:

• The Source clip containing the wire to be removed.

• An optional CleanPlate input to allow you to supply a clean plate. This is used by the Clean Plate repair mode which will warp the clean plate onto the current frame and use the warped image to reconstruct the background behind the wire.

Indicators on the On-Screen Wire Tool

You may see the following indicators appear on the on-screen wire tool:

The current frame is inside the range of frames to track the wire over.

The current frame is not inside the range of frames to track the wire over.

The current frame is the first frame in the range of frames to track the wire over.

The current frame is the last frame in the range of frames to track the wire over.

A user key frame has been set on the current frame.

Tracker Controls The following controls appear in the on-screen tracker panel that you can display in the Viewer by checking Show On Screen Controls in the WireRemoval properties.


WIREREMOVAL 66Tracker Controls

Toggle display mode - Toggle the display mode for the on-screen wire tool: show the points and lines, show just the points, or hide both points and lines.

Number of points - Changes the number of points used to describe the wire.

Create user key frame - Creates a user key frame.

Delete user key frame - Deletes a user key frame.

Snap to wire - Finds the edges of the wire and snaps the edges of the region onto them.

Track backwards - Plays backwards through the sequence tracking from frame to frame.

Step backward - Tracks backwards one frame.

Step forward - Tracks forward one frame.

Track forwards - Plays forwards through the sequence tracking from frame to frame.

Smart track - Tracks from beginning to end of frame range in an intelligent order.

Delete track key frames backwards - Deletes track key frames backwards through the sequence until either a user key frame or the beginning of the sequence is reached.

Delete track key frame and step backward - Deletes a track key frame and steps backwards one frame.


WIREREMOVAL 67Parameters

Delete track key frame - Delete the current track key frame.

Delete track key frame and step forwards - Deletes a track key frame and steps forwards one frame.

Delete track key frames forwards - Deletes track key frames forwards through the sequence until either a user key frame or the end of the sequence is reached.

Delete all track key frames - Deletes all track key frames from the sequence.

Delete all track and user key frames - Deletes both track key frames and user key frames.

The corresponding controls are also available in the WireRemoval properties.


Tracker buttons - most of the on-screen tracker panel buttons also appear here. For more information, see “Tracker Controls” on page 65.

Type - this parameter controls the number of points on the on-screen wire tool. Choose the number of points needed to describe the wire you wish to remove.

• Two Points - choose this if your wire is straight.

• Three Points - choose this if your wire is a simple curve.

• Five Points - choose this if your wire has an s-shaped curve.

On-Screen Wire - the display mode for the on-screen wire tool.

• Show - show both points and lines.

• Hide - hide both points and lines.

• Points only - only show the points.

Show On Screen Controls - use this to show or hide the tracker panel in the Viewer.

Output - sets the output mode for the plug-in.

• Source - output the untouched source image. Use this output mode to position the on-screen wire tool over the wire you wish to remove.



• Repair - output the repaired source image, with the wire removed from under the on-screen tool.

• Wire Matte - renders a matte for the wire. This may be useful if the wire has been tracked but cannot be repaired using F_WireRemoval and other techniques have to be used.

• Repair Matted - output the repaired source image and a matte in the alpha channel. If you want, you can adjust your image further manually using the matte.

Track Range - the range of frames to track the wire over.

• Specified Range - use the Track Start and Track End parameters to specify the range over which to track the wire.

• Source Clip Range - track the wire over the entire range of the Source clip.

Track Start - use this to specify the start of the tracking range when Track Range has been set to Specified Range.

Track End - use this to specify the end of the tracking range when Track Range has been set to Specified Range.

Repair - sets the algorithm used to remove the wire from under the grain.

• Spatial - this method uses a slope dependent filter that interpolates across the wire at the most likely angle, given the image behind the wire. It uses information from the current frame only.

• Temporal With Static Scene - this method uses local motion estimation to align frames from before and after onto the current frame. If the wire is not in the same place with respect to the background in these two frames, it can then use background information from them to fill in the background behind the wire in the current frame. Where no such information is available, for example, if the wire covers part of the same region in all three frames, the spatial repair method above will be used. This is useful for sequences where the wire is moving and where the motion in the rest of the scene is non-uniform, for example, if the camera has been locked to place but there are objects moving in the area surrounding the wire.

• Temporal With Moving Scene - also aligns frames from before and after onto the current frame, but uses global motion estimation. Again, it gets background information from these two frames where it can and uses the spatial repair method to fill in the rest. This is useful for sequences where the wire is moving and the motion in the rest of the scene is fairly uniform, for example, if the entire scene is moving in the same direction as a result of a camera pan.



• Clean Plate - choose this method if you have a clean plate you wish to use for the repair, or if F_WireRemoval does not do a good job of removing the wire from each frame. Connect the clean plate or single frame with the wire removed to the CleanPlate input. F_WireRemoval will then align this frame to the frame to be repaired in order to reconstruct the background behind the wire.

Filter Size - this is a trade off between the amount of grain removed and the blurring of image along the direction of the wire. If the wire you are trying to remove has detail within it (for example, a steel wire in which the twisted threads are reflecting light), then the algorithm may leave these alone, thinking that they are grain. In this situation, you should decrease the filter size. A value of zero will definitely remove the artefacts but also the grain, which would then have to be put back using FurnaceCore’s F_ReGrain.

Temporal Offset - the time offset of the additional frames to use for the Temporal With Static Scene or Temporal With Moving Scene methods. This determines which two frames before and after the current frame are used to fill in the background behind the wire. For example, if this is set to 2 and the current frame is 30, frames 28 and 32 are used.

Luminance Correct - turn this on for methods other than Spatial repair where there are global luminance shifts between one frame of the sequence and the next, or between a frame of the sequence and a clean plate you are using for the repair. With Luminance Correct turned on, F_WireRemoval will adjust the luminance of the background pixels it uses to the correct level before doing the repair.

Lum Block Size - change this value if luminance correction is not working as well as it should. The luminance correction uses overlapping blocks and matches the luminance within those blocks; changing the size of the blocks will change the regions covered by each block and could result in a better match.

Points - Parameters to set the position of the points used to define the wire.

Point 1 - the position of the start point on the wire.

Point 2 - the position of the point on the wire between the start point and the mid point (this is only active if Type is Five Points).

Point 3 - the position of the mid point on the wire.

Point 4 - the position of the point on the wire between the mid point and the end point (this is only active if Type is Five Points).

Point 5 - the position of the end point on the wire.



Start Width - the width of the wire at point 1 of the on-screen wire tool.

End Width - the width of the wire at point 5 of the on-screen wire tool. This allows you to make your repair region wider at one end than the other, which is useful for cases where there is motion blur on the wire.

Overall Width - increase this parameter to expand the width of the repair region along its entire length, and for all key frames.


GLOBAL MOTION ESTIMATION

Introduction FurnaceCore has three effects based on global motion estimation (GME). They all calculate a four-corner pin, which finds the best fit of one image onto another, and then apply that pin. These effects differ in how that corner pin is calculated and to which image that pin is applied. This chapter describes the general idea behind these plug-ins; for more detailed information on each plug-in, please read their individual chapters.

These effects are:

• F_Align - which lines up two shots of the same scene, by finding a corner pin from each source clip frame to the corresponding reference clip frame. For example, you can use this to align two separate but similar steady cam shots of the same scene.

• F_RigRemoval - which removes unwanted objects (rigs) from image sequences without accurate rotoscoping or keying to produce a clean plate.

• F_Steadiness - which removes motion from a single clip, by calculating a pin that either locks all frames in the clip to a single reference frame, or smooths that motion out over a window of frames. For example, you can use this to remove camera shake.

What is Global Motion Estimation?

Global motion estimation is a technique that attempts to map one image onto another with a simple four-corner pin. This differs from local motion estimation (LME), which attempts to find where each individual pixel in the image is in the other image. (For more information about local motion estimation, please see “Local Motion Estimation” on page 76). GME is much cheaper to compute than LME, but gives you less information about the image. Nevertheless, it is still very powerful for a variety of applications.

Using the plug-ins’ parameters, you can tell the GME engine what type of motion to expect. This can be a combination of any of:

1. translation - which allows the four corners to translate by the same amount,

2. rotation - which allows the corners to rotate about their centre,

3. scale - which allows the size of the area defined by the corners to change,

4. perspective - which allows the angles at the corners to change, so that the area defined by them is no longer a rectangle.

GLOBAL MOTION ESTIMATION 72The Analysing Global Motion Estimation Effects

The more types of motion you allow, the more expensive the motion estimation becomes. For many scenes, rotation and translation are sufficient.

The GME effects have an accuracy control, which controls the amount of work The Foundry’s GME engine does to calculate the global motion. Typically, the higher this is, the better the estimation, but the more expensive it is.

Limitations of GME As stated above, global motion estimation simply calculates a four-corner pin to transform one image onto another. This means that GME can’t be used to match two images where there is heavy parallax, very complicated foreground motion, changing objects, and so on.

The best way to think of what GME can do is that if you can do it with a four-corner pin, it can; if you can’t, it can’t. However, GME will take the pain out of hand matching pins frame by frame.

The Analysing Global Motion Estimation Effects

F_Align and F_Steadiness work in a similar way, which is distinct from the way F_RigRemoval works. These two effects calculate a four-corner pin for each frame and save it into the corner pin parameters. These pins are then used during the render to move the source image.

Using Them These effects analyse images over a range of frames to figure out their four-corner pins. This is done in response to the user pressing the Analyse button in the effects control panel. During analysis, the effect will run through a range of frames adding keys to the corner pin parameters. These corner pins are then applied to the source clip to render a frame.

F_Steadiness has a separate analysis pass that happens interactively; it then uses the previously computed and key-framed corner pins during render. This speeds up its operation, as the analysis step only has to be done once, and once it has been done this plug-in is very quick to render. However, F_Align, which only ever needs the two current frames from each clip, can compute the corner pin on the fly (but not keyframe it!). This leads to a slightly different mode of operation for the following effects:

• F_Steadiness

• needs to have an analysis run before it renders useful output,

• will always use the value in the corner pin parameters when rendering the output image.


GLOBAL MOTION ESTIMATION 73Controls

• F_Align

• no need to have the analysis run to render output, but doing so will give you a key-framed corner pin,

• during render, it will use the value of the corner pin parameters only if there is a keyframe there, otherwise, it will analyse on the fly during render. This means that analysis will speed up later renders, as just rendering the corner pin is much cheaper than calculating it.

Some parameters to the effect control how the effect performs GME analysis, and some only affect the rendering. If you ever modify one of these parameters, then any analysis you may have performed will be out of date. To let you know, an overlay warning will be posted whenever this happens. You don’t have to re-analyse and your renders will still look at the keyed corner pins.

If you have not modified a parameter that affects analysis (the warning overlay will let you know), pressing Analyse will only re-analyse a frame if there is no key on the corner pin at that time. This avoids redundant re-analysis if you have interrupted the analysis or extended the analysis range. However, if you want to force re-analysis, press Clear Analysis and all keys will be deleted.

F_Steadiness and F_Align have an analysis region rectangle parameter which is used to specify which area of the reference image should be analysed during GME. So, for example, with F_Steadiness set to Lock Mode, this is the area inside the lock frame that a match will be sought for. The documentation for each plug-in describes exactly how to use the analysis region.

Controls The controls common to all GME plug-ins are described below. They are grouped into two sections: ones that determine how analysis is carried out, and ones that control the rendering of the output.

Controls That Affect Analysis

The following parameters and controls affect the analysis of the four-corner pin.

Analyse - This is a push button which will trigger an analysis of the input clips and calculate a corner pin. Interrupting the analysis will not delete the corner pin keys that have already been calculated.


GLOBAL MOTION ESTIMATION 74Controls

Render During Analysis - If set, this toggle will cause the effect to update the time line and render a freshly analysed frame so you can see the progress of the effect. Doing so will slow down the analysis somewhat, so toggle this off to speed up the general analysis.

Clear Analysis - Pressing this push button will delete all keyframes from the corner pin parameters, allowing you to force a re-analysis if you feel the need to.













Accuracy - This controls the time/accuracy trade off in the GME engine. The


GLOBAL MOTION ESTIMATION 75Widgets

higher this is, the slower it goes, but you have a better likelihood of a good result.

Parameters That Affect Rendering

These following parameters control how a GME effect renders the four-corner pin. Some of them are set during the analysis pass.

Filtering - This controls the quality of the rendering.

• Low - uses nearest neighbour filtering. This gives low quality but is quick to render.



Invert - If set, then the inverse of the calculated four-corner pin is used during render.


• Bottom Left - the lower left corner pin.

• Bottom Right - the lower right corner pin.

• Top Left - the upper left corner pin.

• Top Right - the upper right corner pin.

Widgets All the Analysing GME effects have two on-screen widgets: one to provide feedback and one to set up the analysis region.

Analysis Region Widget - This is a rectangle widget which you use to set the analysis region over the reference image.

Four Corner Widget - This is a widget that shows the state of the four-corner pin that has been calculated. You can change it by grabbing any of the corners and tweaking the shape of the pin. To give you more feedback as to what frames have been analysed, it will be drawn solid if there is a key in the corner pin at the frame being displayed; otherwise, it will be drawn dashed.


LOCAL MOTION ESTIMATION

Introduction Several tools in FurnaceCore make use of Motion Estimation technology. Motion Estimation tends to fall into two areas: Global Motion Estimation and Local Motion Estimation. In this chapter, we look at the parameters for Local Motion Estimation (or LME), which is the per-pixel motion analysis used in tools such as F_Kronos.

Background The easiest way to understand LME is to think in terms of vector fields. A vector field for an image in a sequence has the same dimensions as the image, but contains an (x,y) offset per pixel. These offsets show how to warp a neighbouring image onto the current image. Clearly, as most of the images in a sequence have two neighbours, each can have two vector fields. These are called the ’forward motion vectors’ where they represent the warp of the image in front of the current one, and ’backward motion vectors’ where they represent the warp of the image behind the current one.

This is an approximation to what is really going on in an image sequence. A single vector field can be thought of as representing a warp or a morph - sometimes referred to as a ’rubber sheet’ warp and cannot truly represent multiple overlapping motions. This effect can be seen where moving foreground objects appear to drag the background. To help cope with this restriction, a number of the tools allow the use of two vector fields per frame, one representing foreground motion and one representing background motion, where a matte input is used to identify the separation. In addition, see the Occlusions option in the parameter descriptions below, which attempts to improve the rubber-sheet effect when separate vector fields aren’t being used.

LME is a computationally expensive operation; most of the tools which use LME have a choice between generating vector fields on the fly and using pre-calculated vector fields. When using pre-calculated vector fields, these are typically generated by the F_VectorGenerator tool (described on page 56), although they may also come from other third-party sources.

The vector generation process has a number of tuning parameters which can be used to adapt the vectors to suit particular sequences, as well as to trade off render time verses accuracy of vectors.

LOCAL MOTION ESTIMATION 77Using Pre-Calculated Vector Fields

Using Pre-Calculated Vector Fields

LME is a computationally expensive operation. For this reason, most of the plug-ins which use LME allow you to choose between generating vector fields on the fly and using pre-calculated vector fields. Pre-calculated vector fields can be generated by the F_VectorGenerator tool, described on page 56, and may also come from third-party sources. These can then be passed into the following effects as inputs:

• F_Kronos

• F_MotionBlur

Therefore, if you are going to be using more than one of these effects in your project, it might be worth generating the vector fields beforehand with F_VectorGenerator, so that they can be reused.

Most of the time, a vector field that produces good output in one of these effects will work well in the others as well.

Parameters Not all tools using LME have all of these parameters. They can be categorised into two groups: one for the generation of the vectors, and one for their use in picture warping.

Vector Generation Parameters

Vector Detail - Adjust this to vary the resolution of the vector field. The larger vector detail is, the greater the processing time, but the more detailed the vectors should be. A value of 1.0 will generate a vector at each pixel. A value of 0.5 will generate a vector at every other pixel. For some sequences, a high vector detail near 1.0 generates too much unwanted local motion detail, and often a low value is more appropriate.

Smoothness - Vector fields usually have two important qualities: they should accurately match similar pixels in one image to another and they should be smooth rather than noisy. Often, it is necessary to trade one of these qualities off against the other. A high smoothness will miss lots of local detail, but is less likely to provide you with the odd spurious vector (think of it as if the vectors were combed until they are smooth with respect to their neighbours). A low smoothness will concentrate on detail matching, even if the resulting field is jagged. The default value of 0.5 should work well for most sequences.

Block Size - To find where a certain pixel is located in the analysis range, the vector generation algorithm looks for a block of pixels centered around that pixel. Block size defines the width and height of these blocks (in pixels). On rare occasions, a large block size can produce data that’s lacking in


LOCAL MOTION ESTIMATION 78Parameters

detail. This is because a small feature can fit into a large block, causing the motion estimation to concentrate on the background motion and ignore the small feature. A small value, instead, can produce a noisy motion field, as there aren’t enough constraints in a small block to fit the motion accurately. In most cases, however, the default value is small enough so that details aren’t lost, and the smoothing step of the algorithm ensures the motion field isn’t too noisy. Therefore, this value very rarely needs editing.

Over Smooth - This is a computationally intensive smoothing operation that performs a different vector-smoothing operation to normal. This generates highly smooth vector fields (as if the vectors were combed several times with a fine-toothed comb), but may also sacrifice a lot of required detail. In most cases, this level of smoothing isn’t necessary.

Luminance Correct - LME is highly dependent upon the idea that the brightness of objects doesn’t vary through a sequence. Where brightness varies rapidly - for example a highlight moving across the bodywork of a car - the motion calculation will perform poorly. The luminance of a shot can come from other sources too - such as an overall flicker problem. In these cases where there is a global luminance shift, toggling this control on will allow the LME algorithm to take account of overall brightness changes between frames.

Tolerances - For efficiency, much of the LME is done on luminance only - i.e., using monochrome images. The tolerances allow you to tune the weight of each colour channel when calculating the image luminance. These parameters rarely need tuning. However, you may, for example, wish to increase the red weighting Weight Red to allow the algorithm to concentrate on getting the motion of a primarily red object correct, at the cost of the rest of the items in a shot.

Weight Red

Weight Green

Weight Blue

Picture Warping Parameters

Filtering - Sets the quality of filtering when producing in-between (F_Kronos) frames.

• Normal - use bilinear interpolation which gives good results and is a lot quicker than Extreme.

• Extreme - uses a sinc interpolation filter to give a sharper picture but takes a lot longer to render.

Warp Mode - Sets how to control the new timing of the clip.


LOCAL MOTION ESTIMATION 79Vector Field Representation

• Simple - this is the quickest option, but may produce less than optimal results around moving objects and image edges.

• Normal - this is the standard option, with more optimal treatment of moving objects and image edges.

• Occlusions - this is an advanced option that can improve the results when not doing a separated picture build with multiple vector fields and mattes. It attempts to reduce the level of background dragging that occurs between foreground and background objects.

Show Vectors - Switch this on to display the vectors on the screen.

Vector Field Representation

Digital images generally consist of the four standard channels: red, green, blue, and alpha. Previous versions of Furnace, in hosts such as Shake, were forced to encode the motion vectors in these chromatic image channels. NukeX, however, allows users and plug-ins to create or grab data from up to 1023 channels.

These additional channels enable the FurnaceCore tools that consume vectors to read from motion vector specific channels called 'forward' and 'backward'. Likewise, the FurnaceCore tools that produce vectors can write to vector specific channels instead of destroying the pre-existing 'rgba' data.

When viewing 'forward' or 'backward' motion vectors, NukeX represents the x values as amounts of red and y values as amounts of green. Motion vectors can be positive or negative, where zero represents no motion.

The 'backward' channel represents the offsets required to fetch pixels from the previous frame to match pixels in the current frame. The 'forward' channel represents the offsets required to fetch pixels from the following frame to match pixels in the current frame.

Figure 61 shows an example of a vector image for the taxi clip shown in Figure 60. These vectors were produced using the FurnaceCore plug-in F_VectorGenerator - for more details, please see “VectorGenerator” on page 56.


LOCAL MOTION ESTIMATION 80Vector Field Representation

Figure 60. Taxi. Figure 61. Motion vectors for the taxi sequence.


APPENDIX A

Third Party Licenses

This appendix lists third party libraries used in FurnaceCore, along with their licenses.

Library Description License

Boost Source code func-tion / template library

Boost Software License - Version 1.0 - August 17th, 2003

Permission is hereby granted, free of charge, to any person or organization obtaining a copy of the software and accompanying documentation covered by this license (the “Soft-ware”) to use, reproduce, display, distribute, execute, and transmit the Software, and to prepare derivative works of the Software, and to permit third-parties to whom the Soft-ware is furnished to do so, all subject to the following:

The copyright notices in the Software and this entire statement, including the above license grant, this restriction and the following disclaimer, must be included in all copies of the Software, in whole or in part, and all derivative works of the Software, unless such copies or derivative works are solely in the form of machine-executable object code generated by a source language processor.

THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FIT-NESS FOR A PARTICULAR PURPOSE, TITLE AND NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR ANYONE DISTRIBUTING THE SOFTWARE BE LIABLE FOR ANY DAMAGES OR OTHER LIABILITY, WHETHER IN CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEAL-INGS IN THE SOFTWARE.

Expat XML parser Copyright © 1998, 1999, 2000 Thai Open Source Software Center Ltd and Clark Cooper

Copyright © 2001, 2002, 2003, 2004, 2005, 2006 Expat maintainers.

Permission is hereby granted, free of charge, to any person obtaining a copy of this soft-ware and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, dis-tribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or sub-stantial portions of the Software.

THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FIT-NESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIA-BILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

APPENDIX A 82Third Party Licenses

FFTW Fast Fourier trans-form library

Copyright © 2003 Matteo Frigo

Copyright © 2003 Massachusetts Institute of Technology

All rights reserved. Used under terms of a commercial license. http://www.fftw.org.

FreeType Font support Portions of this software are copyright © 2008 The FreeType Project

(www.freetype.org). All rights reserved.

FTGL OpenGL support FTGL - OpenGL font library

Copyright © 2001-2004 Henry Maddocks [email protected]

Copyright © 2008 Sam Hocevar [email protected]

Copyright © 2008 Sean Morrison [email protected]

Permission is hereby granted, free of charge, to any person obtaining a copy of this soft-ware and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, dis-tribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions

The above copyright notice and this permission notice shall be included in all copies or sub-stantial portions of the Software.

THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FIT-NESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIA-BILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.



APPENDIX A 83Third Party Licenses

OFX Plug-in API Copyright (c) 2003-2009, The Open Effects Association Ltd. All rights reserved.

Redistribution and use in source and binary forms, with or without modification, are per-mitted provided that the following conditions are met:

• Redistributions of source code must retain the above copyright notice, this list of con-ditions and the following disclaimer.

• Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.

• Neither the name The Open Effects Association Ltd, nor the names of its contributors may be used to endorse or promote products derived from this software without spe-cific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR NY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUEN-TIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.



APPENDIX B

End User Licensing Agreement (EULA)

IMPORTANT: BY INSTALLING THIS SOFTWARE YOU ACKNOWLEDGE THAT YOU HAVE READ THIS AGREEMENT, UNDERSTAND IT AND AGREE TO BE BOUND BY ITS TERMS AND CONDITIONS. IF YOU DO NOT AGREE TO THE TERMS OF THIS AGREEMENT DO NOT INSTALL, COPY OR USE THE SOFTWARE.

This END USER LICENSE AGREEMENT (this "Agreement") is made by and between The Foundry Visionmongers Ltd., a company registered in England and Wales, ("The Foundry"), and you, as either an individual or a single entity ("Licensee").

In consideration of the mutual covenants contained herein and for other good and valuable consideration (the receipt and sufficiency of which is acknowledged by each party hereto) the parties agree as follows:

SECTION 1. GRANT OF LICENSE.

Subject to the limitations of Section 2, The Foundry hereby grants to Licensee a limited, non-transferable and non-exclusive license to install and use a machine readable, object code version of this software program (the "Software") and accompanying user guide and other documentation (collectively, the "Documentation") solely for Licensee's own internal business purposes (collectively, the "License"); provided, however, Licensee's right to install and use the Software and the Documentation is limited to those rights expressly set out in this Agreement.

SECTION 2. RESTRICTIONS ON USE.

Licensee is authorized to use the Software in machine readable, object code form only, and Licensee shall not: (a) assign, sublicense, sell, distribute, transfer, pledge, lease, rent, share or export the Software, the Documentation or Licensee's rights hereunder; (b) alter or circumvent the copy protection mechanisms in the Software or reverse engineer, decompile, disassemble or otherwise attempt to discover the source code of the Software; (c) modify, adapt, translate or create derivative works based on the Software or Documentation; (d) use, or allow the use of, the Software or Documentation on any project other than a project produced by Licensee (an "Authorized Project"); (e) allow or permit anyone (other than Licensee and Licensee's authorized employees to the extent they are working on an Authorized Project) to use or have access to the Software or Documentation; (f) copy or install the Software or Documentation other than as expressly provided for herein; or (g) take any action, or fail to take action, that could adversely affect the trademarks, service marks, patents, trade secrets, copyrights or other intellectual property rights of The Foundry or any third party with intellectual property rights in the Software (each, a "Third Party Licensor"). Furthermore, for purposes of this Section 2, the term "Software" shall include any derivatives of the Software.

Licensee shall install and use only a single copy of the Software on one computer, unless the Software is installed in a "floating license" environment, in which case Licensee may install the Software on more than

APPENDIX B 85End User Licensing Agreement (EULA)

one computer; provided, however, Licensee shall not at any one time use more copies of the Software than the total number of valid Software licenses purchased by Licensee.

Please note that in order to guard against unlicensed use of the Software a licence key is required to access and enable the Software. The issuing of replacement or substituted licence keys if the Software is moved from one computer to another is subject to and strictly in accordance with The Foundry’s Licence Transfer Policy, which is available on The Foundry’s website and which requires a fee to be paid in certain circumstances. The Foundry may from time to time and at its sole discretion vary the terms and conditions of the Licence Transfer Policy.

Furthermore, if the Software can be licensed on an "interactive" or "non-interactive" basis, licensee shall be authorized to use a non-interactive version of the Software for rendering purposes only (i.e., on a CPU, without a user, in a non-interactive capacity) and shall not use such Software on workstations or otherwise in a user-interactive capacity. Licensee shall be authorized to use an interactive version of the Software for both interactive and non-interactive rendering purposes, if available.

If Licensee has purchased the Software on the discount terms offered by The Foundry’s Educational Policy published on its website (“the Educational Policy”), Licensee warrants and represents to The Foundry as a condition of this Agreement that: (a) (if Licensee is an individual) he or she is a part-time or full-time student at the time of purchase and will not use the Software for commercial, professional or for-profit purposes; (b) (if the Licensee is not an individual) it is an organisation that will use it only for the purpose of training and instruction, and for no other purpose (c) Licensee will at all times comply with the Educational Policy (as such policy may be amended from time to time).

Finally, if the Software is a "Personal Learning Edition," (“PLE”) Licensee may use it only for the purpose of personal or internal training and instruction, and for no other purpose. PLE versions of the Software may not be used for commercial, professional or for-profit purposes including, for the avoidance of doubt, the purpose of providing training or instruction to third parties.

SECTION 3. SOURCE CODE.

Notwithstanding that Section 1 defines “Software” as an object code version and that Section 2 provides that Licensee may use the Software in object code form only, The Foundry may also agree to license to Licensee (including by way of upgrades, updates or enhancements) source code or elements of the source code of the Software the intellectual property rights in which belong either to The Foundry or to a Third Party Licensor (“Source Code”). If The Foundry does so Licensee shall be licensed to use the Source Code as Software on the terms of this Agreement and: (a) notwithstanding Section 2 (c) Licensee may use the Source Code at its own risk in any reasonable way for the limited purpose of enhancing its use of the Software solely for its own internal business purposes and in all respects in accordance with this Agreement; (b) Licensee shall in respect of the Source Code comply strictly with all other restrictions applying to its use of the Software under this Agreement as well as any other restriction or instruction that is communicated to it by The Foundry at any time during this Agreement (whether imposed or requested by The Foundry or by any Third Party Licensor); (c) notwithstanding any other term of this Agreement The Foundry gives no warranty whatsoever in respect of the Source Code, which is licensed on an “as is” basis, or in respect of any modification of the Source Code made by Licensee (“Modification”); (d)



notwithstanding any other term of this Agreement The Foundry shall have no obligation to provide support, maintenance, upgrades or updates of or in respect of the Source Code or of any Modification; and (e) Licensee shall indemnify The Foundry against all liabilities and expenses (including reasonable legal costs) incurred by The Foundry in relation to any claim asserting that any Modification infringes the intellectual property rights of any third party.

SECTION 4. BACK-UP COPY.

Notwithstanding Section 2, Licensee may store one copy of the Software and Documentation off-line and off-site in a secured location owned or leased by Licensee in order to provide a back-up in the event of destruction by fire, flood, acts of war, acts of nature, vandalism or other incident. In no event may Licensee use the back-up copy of the Software or Documentation to circumvent the usage or other limitations set forth in this Agreement.

SECTION 5. OWNERSHIP.

Licensee acknowledges that the Software (including, for the avoidance of doubt, any Source Code that is licensed to Licensee) and Documentation and all intellectual property rights and other proprietary rights relating thereto are and shall remain the sole property of The Foundry and the Third Party Licensors. Licensee shall not remove, or allow the removal of, any copyright or other proprietary rights notice included in and on the Software or Documentation or take any other action that could adversely affect the property rights of The Foundry or any Third Party Licensor. To the extent that Licensee is authorized to make copies of the Software or Documentation under this Agreement, Licensee shall reproduce in and on all such copies any copyright and/or other proprietary rights notices provided in and on the materials supplied by The Foundry hereunder. Nothing in this Agreement shall be deemed to give Licensee any rights in the trademarks, service marks, patents, trade secrets, confidential information, copyrights or other intellectual property rights of The Foundry or any Third Party Licensor, and Licensee shall be strictly prohibited from using the name, trademarks or service marks of The Foundry or any Third Party Licensor in Licensee's promotion or publicity without The Foundry's express written approval.

SECTION 6. LICENSE FEE.

Licensee understands that the benefits granted to Licensee hereunder are contingent upon Licensee's payment in full of the license fee payable in connection herewith (the "License Fee").

SECTION 7. UPGRADES/ENHANCEMENTS.

The Licensee's access to support, upgrades and updates is subject to the terms and conditions of the "Annual Upgrade and Support Programme” available on The Foundry's website. The Foundry may from time to time and at its sole discretion vary the terms and conditions of the Annual Upgrade and Support Programme.

SECTION 8. TAXES AND DUTIES.

Licensee agrees to pay, and indemnify The Foundry from claims for, any local, state or national tax (exclusive of taxes based on net income), duty, tariff or other impost related to or arising from the transaction contemplated by this Agreement.



SECTION 9. LIMITED WARRANTY.

The Foundry warrants that, for a period of ninety (90) days after delivery of the Software: (a) the machine readable electronic files constituting the Software and Documentation shall be free from errors that may arise from the electronic file transfer from The Foundry and/or its authorized reseller to Licensee; and (b) to the best of The Foundry's knowledge, Licensee's use of the Software in accordance with the Documentation will not, in and of itself, infringe any third party's copyright, patent or other intellectual property rights. Except as warranted, the Software and Documentation is being provided "as is." THE FOREGOING LIMITED WARRANTY IS IN LIEU OF ALL OTHER WARRANTIES OR CONDITIONS, EXPRESS OR IMPLIED, AND The Foundry DISCLAIMS ANY AND ALL IMPLIED WARRANTIES OR CONDITIONS, INCLUDING, WITHOUT LIMITATION, ANY IMPLIED WARRANTY OF TITLE, NON-INFRINGEMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, REGARDLESS OF WHETHER The Foundry KNOWS OR HAS REASON TO KNOW OF LICENSEE'S PARTICULAR NEEDS. The Foundry does not warrant that the Software or Documentation will meet Licensee's requirements or that Licensee's use of the Software will be uninterrupted or error free. No employee or agent of The Foundry is authorized to modify this limited warranty, nor to make additional warranties. No action for any breach of the above limited warranty may be commenced more than one (1) year after Licensee's initial receipt of the Software. To the extent any implied warranties may not be disclaimed under applicable law, then ANY IMPLIED WARRANTIES ARE LIMITED IN DURATION TO NINETY (90) DAYS AFTER DELIVERY OF THE SOFTWARE TO LICENSEE.

SECTION 10. LIMITED REMEDY.

The exclusive remedy available to the Licensee in the event of a breach of the foregoing limited warranty, TO THE EXCLUSION OF ALL OTHER REMEDIES, is for Licensee to destroy all copies of the Software, send The Foundry a written certification of such destruction and, upon The Foundry's receipt of such certification, The Foundry will make a replacement copy of the Software available to Licensee.

SECTION 11. INDEMNIFICATION.

Licensee agrees to indemnify, hold harmless and defend The Foundry, the Third Party Licensors and The Foundry's and each Third Party Licensor’s respective affiliates, officers, directors, shareholders, employees, authorized resellers, agents and other representatives (collectively, the "Released Parties") from all claims, defense costs (including, but not limited to, attorneys' fees), judgments, settlements and other expenses arising from or connected with the operation of Licensee's business or Licensee's possession or use of the Software or Documentation.

SECTION 12. LIMITED LIABILITY.

In no event shall the Released Parties' cumulative liability to Licensee or any other party for any loss or damages resulting from any claims, demands or actions arising out of or relating to this Agreement (or the Software or Documentation contemplated herein) exceed the License Fee paid to The Foundry or its authorized reseller for use of the Software. Furthermore, IN NO EVENT SHALL THE RELEASED PARTIES BE LIABLE TO LICENSEE UNDER ANY THEORY FOR ANY INDIRECT, SPECIAL, INCIDENTAL, PUNITIVE, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING DAMAGES FOR LOSS OF BUSINESS OR LOSS OF PROFITS) OR THE COST OF PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES, REGARDLESS OF WHETHER THE RELEASED PARTIES KNOW OR HAVE REASON TO KNOW OF THE POSSIBILITY OF SUCH DAMAGES AND REGARDLESS OF WHETHER ANY REMEDY SET FORTH HEREIN FAILS OF ITS ESSENTIAL



PURPOSE. No action arising out of or related to this Agreement, regardless of form, may be brought by Licensee more than one (1) year after Licensee's initial receipt of the Software; provided, however, to the extent such one (1) year limit may not be valid under applicable law, then such period shall be limited to the shortest period allowed by law.

SECTION 13. TERM; TERMINATION.

This Agreement is effective upon Licensee's acceptance of the terms hereof and Licensee's payment of the License Fee, and the Agreement will remain in effect until termination. If Licensee breaches this Agreement, The Foundry may terminate the License granted hereunder by notice to Licensee. In the event the License is terminated, Licensee will either return to The Foundry all copies of the Software and Documentation in Licensee's possession or, if The Foundry directs in writing, destroy all such copies. In the later case, if requested by The Foundry, Licensee shall provide The Foundry with a certificate signed by an officer of Licensee confirming that the foregoing destruction has been completed.

SECTION 14. CONFIDENTIALITY.

Licensee agrees that the Software (including, for the avoidance of doubt, any Source Code that is licensed to Licensee) and Documentation are proprietary and confidential information of The Foundry or, as the case may be, the Third Party Licensors, and that all such information and any communications relating thereto (collectively, "Confidential Information") are confidential and a fundamental and important trade secret of The Foundry or the Third Party Licensors. Licensee shall disclose Confidential Information only to Licensee's employees who are working on an Authorized Project and have a "need-to-know" of such Confidential Information, and shall advise any recipients of Confidential Information that it is to be used only as authorized in this Agreement. Licensee shall not disclose Confidential Information or otherwise make any Confidential Information available to any other of the Licensee's employees or to any third parties without the express written consent of The Foundry. Licensee agrees to segregate, to the extent it can be reasonably done, the Confidential Information from the confidential information and materials of others in order to prevent commingling. Licensee shall take reasonable security measures, which such measures shall be at least as great as the measures Licensee uses to keep Licensee's own confidential information secure (but in any case using no less than a reasonable degree of care), to hold the Software, Documentation and any other Confidential Information in strict confidence and safe custody. The Foundry may request, in which case Licensee agrees to comply with, certain reasonable security measures as part of the use of the Software and Documentation. Licensee acknowledges that monetary damages may not be a sufficient remedy for unauthorized disclosure of Confidential Information, and that The Foundry shall be entitled, without waiving any other rights or remedies, to such injunctive or equitable relief as may be deemed proper by a court of competent jurisdiction.

SECTION 15. INSPECTION.

Licensee shall advise The Foundry on demand of all locations where the Software or Documentation is used or stored. Licensee shall permit The Foundry or its authorized agents to inspect all such locations during normal business hours and on reasonable advance notice.



SECTION 16. NONSOLICITATION.

Licensee agrees not to solicit for employment or retention any of The Foundry's current or future employees who were or are involved in the development and/or creation of the Software.

SECTION 17. U.S. GOVERNMENT LICENSE RIGHTS.

The Software, Documentation and/or data delivered hereunder are subject to the terms of this Agreement and in no event shall the U.S. Government acquire greater than RESTRICTED/LIMITED RIGHTS. At a minimum, use, duplication or disclosure by the U.S. Government is subject to the applicable restrictions of: (i) FAR §52.227-14 ALTS I, II and III (June 1987); (ii) FAR §52.227-19 (June 1987); (iii) FAR §12.211 and 12.212; and/or (iv) DFARS §227.7202-1(a) and DFARS §227.7202-3.

The Software is the subject of the following notices:

• Copyright (c) 2/16/12 The Foundry Visionmongers, Ltd.. All Rights Reserved.

• Unpublished-rights reserved under the Copyright Laws of the United Kingdom.

SECTION 18. SURVIVAL.Sections 2, 3, 5, 6, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 and 20 shall survive any termination or expiration of this Agreement.

SECTION 19. IMPORT/EXPORT CONTROLS.To the extent that any Software made available hereunder is subject to restrictions upon export and/or reexport from the United States, Licensee agrees to comply with, and not act or fail to act in any way that would violate, the applicable international, national, state, regional and local laws and regulations, including, without limitation, the United States Foreign Corrupt Practices Act, the Export Administration Act and the Export Administration Regulations, as amended or otherwise modified from time to time, and neither The Foundry nor Licensee shall be required under this Agreement to act or fail to act in any way which it believes in good faith will violate any such laws or regulations.

SECTION 20. MISCELLANEOUS.This Agreement is the exclusive agreement between the parties concerning the subject matter hereof and supersedes any and all prior oral or written agreements, negotiations, or other dealings between the parties concerning such subject. This Agreement may be modified only by a written instrument signed by both parties. If any action is brought by either party to this Agreement against the other party regarding the subject matter hereof, the prevailing party shall be entitled to recover, in addition to any other relief granted, reasonable attorneys' fees and expenses of litigation. Should any term of this Agreement be declared void or unenforceable by any court of competent jurisdiction, such declaration shall have no effect on the remaining terms of this Agreement. The failure of either party to enforce any rights granted hereunder or to take action against the other party in the event of any breach hereunder shall not be deemed a waiver by that party as to subsequent enforcement of rights or subsequent actions in the event of future breaches. This Agreement shall be governed by, and construed in accordance with English Law.

The Foundry and Licensee intend that each Third Party Licensor may enforce against Licensee under the Contracts (Rights of Third Parties) Act 1999 ("the Act") any obligation owed by Licensee to The Foundry



under this Agreement that is capable of application to any proprietary or other right of that Third Party Licensor in or in relation to the Software. The Foundry and Licensee reserve the right under section 2(3)(a) of the Act to rescind, terminate or vary this Agreement without the consent of any Third Party Licensor.

Copyright (c) 2/16/12 The Foundry Visionmongers Ltd. All Rights Reserved. Do not duplicate.


INDEX

A-Z

Aadding

grain 35motion blur 31

Align 13aligning

clips spatially 13alignment tools 12analysing global motion estimation

effects 72

Bbackward vector fields 56, 76

Ccamera motion

removing 50stabilising 50

colour matching 28contacting The Foundry 8contexts 8customer support 8

DDeFlicker2 19

Eend user license agreement 84

Ffilm grain

adding 35stocks 37

flickerreducing 19

forward vector fields 56, 76Furnace Core 6

Ggenerating

motion vectors 56global motion effects 71global motion estimation 71

limitations 72

GME 71grading tools 12grain

adding 35stocks 37

grain management tools 10

Iinstall directory 6installation 6introduction 6

KKronos 21

Llicensing 7limitations

of global motion estimation 72LME 76local motion estimation 76

MMatchGrade 28motion

stabilising 50motion blur

adding 31motion estimation

global 71local 76

motion vector inputs 7motion vectors 7

generating 56precalculated 77pre-calculating 56

MotionBlur 31

Oocclusions in rig removal 46OFX plug-ins 6on-screen tools 7, 63overview

of the plug-ins 10

Pplug-in contexts 8

proxy resolution warnings 39

RReGrain 35removing

camera motion 50flicker 19rigs 43, 71unwanted objects 11, 43, 71wires 60

retiming 21retiming tools 11rig removal 43, 71RigRemoval 43

Sslowing down footage 21spatial alignment of clips 13speeding up footage 21stabilisation tools 12stabilising footage 50Steadiness 50

TThe Foundry

contacting 8The Foundry products 8

Uunwanted objects

removing 43, 71

Vvector channels 57vector fields 76

backward 56, 76forward 56, 76precalculated 77pre-calculating 56representation 79

VectorGenerator 56

Wweb site 8WireRemoval 60www.thefoundry.co.uk 8

Furnace Core

Documents

separated

odd spurious

select furnace

intellectual

global motion

global luminance

sole discretion

local motion